dev.c 154.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9
/*
 * 	NET3	Protocol independent device support routines.
 *
 *		This program is free software; you can redistribute it and/or
 *		modify it under the terms of the GNU General Public License
 *		as published by the Free Software Foundation; either version
 *		2 of the License, or (at your option) any later version.
 *
 *	Derived from the non IP parts of dev.c 1.0.19
10
 * 		Authors:	Ross Biro
L
Linus Torvalds 已提交
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76
 *				Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 *				Mark Evans, <evansmp@uhura.aston.ac.uk>
 *
 *	Additional Authors:
 *		Florian la Roche <rzsfl@rz.uni-sb.de>
 *		Alan Cox <gw4pts@gw4pts.ampr.org>
 *		David Hinds <dahinds@users.sourceforge.net>
 *		Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
 *		Adam Sulmicki <adam@cfar.umd.edu>
 *              Pekka Riikonen <priikone@poesidon.pspt.fi>
 *
 *	Changes:
 *              D.J. Barrow     :       Fixed bug where dev->refcnt gets set
 *              			to 2 if register_netdev gets called
 *              			before net_dev_init & also removed a
 *              			few lines of code in the process.
 *		Alan Cox	:	device private ioctl copies fields back.
 *		Alan Cox	:	Transmit queue code does relevant
 *					stunts to keep the queue safe.
 *		Alan Cox	:	Fixed double lock.
 *		Alan Cox	:	Fixed promisc NULL pointer trap
 *		????????	:	Support the full private ioctl range
 *		Alan Cox	:	Moved ioctl permission check into
 *					drivers
 *		Tim Kordas	:	SIOCADDMULTI/SIOCDELMULTI
 *		Alan Cox	:	100 backlog just doesn't cut it when
 *					you start doing multicast video 8)
 *		Alan Cox	:	Rewrote net_bh and list manager.
 *		Alan Cox	: 	Fix ETH_P_ALL echoback lengths.
 *		Alan Cox	:	Took out transmit every packet pass
 *					Saved a few bytes in the ioctl handler
 *		Alan Cox	:	Network driver sets packet type before
 *					calling netif_rx. Saves a function
 *					call a packet.
 *		Alan Cox	:	Hashed net_bh()
 *		Richard Kooijman:	Timestamp fixes.
 *		Alan Cox	:	Wrong field in SIOCGIFDSTADDR
 *		Alan Cox	:	Device lock protection.
 *		Alan Cox	: 	Fixed nasty side effect of device close
 *					changes.
 *		Rudi Cilibrasi	:	Pass the right thing to
 *					set_mac_address()
 *		Dave Miller	:	32bit quantity for the device lock to
 *					make it work out on a Sparc.
 *		Bjorn Ekwall	:	Added KERNELD hack.
 *		Alan Cox	:	Cleaned up the backlog initialise.
 *		Craig Metz	:	SIOCGIFCONF fix if space for under
 *					1 device.
 *	    Thomas Bogendoerfer :	Return ENODEV for dev_open, if there
 *					is no device open function.
 *		Andi Kleen	:	Fix error reporting for SIOCGIFCONF
 *	    Michael Chastain	:	Fix signed/unsigned for SIOCGIFCONF
 *		Cyrus Durgin	:	Cleaned for KMOD
 *		Adam Sulmicki   :	Bug Fix : Network Device Unload
 *					A network device unload needs to purge
 *					the backlog queue.
 *	Paul Rusty Russell	:	SIOCSIFNAME
 *              Pekka Riikonen  :	Netdev boot-time settings code
 *              Andrew Morton   :       Make unregister_netdevice wait
 *              			indefinitely on dev->refcnt
 * 		J Hadi Salim	:	- Backlog queue sampling
 *				        - netif_rx() feedback
 */

#include <asm/uaccess.h>
#include <linux/bitops.h>
77
#include <linux/capability.h>
L
Linus Torvalds 已提交
78 79 80
#include <linux/cpu.h>
#include <linux/types.h>
#include <linux/kernel.h>
81
#include <linux/hash.h>
82
#include <linux/slab.h>
L
Linus Torvalds 已提交
83
#include <linux/sched.h>
A
Arjan van de Ven 已提交
84
#include <linux/mutex.h>
L
Linus Torvalds 已提交
85 86 87 88 89 90 91 92 93
#include <linux/string.h>
#include <linux/mm.h>
#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/errno.h>
#include <linux/interrupt.h>
#include <linux/if_ether.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
94
#include <linux/ethtool.h>
L
Linus Torvalds 已提交
95 96
#include <linux/notifier.h>
#include <linux/skbuff.h>
97
#include <net/net_namespace.h>
L
Linus Torvalds 已提交
98 99 100 101 102 103
#include <net/sock.h>
#include <linux/rtnetlink.h>
#include <linux/stat.h>
#include <net/dst.h>
#include <net/pkt_sched.h>
#include <net/checksum.h>
104
#include <net/xfrm.h>
L
Linus Torvalds 已提交
105 106 107 108 109 110 111 112
#include <linux/highmem.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/netpoll.h>
#include <linux/rcupdate.h>
#include <linux/delay.h>
#include <net/iw_handler.h>
#include <asm/current.h>
S
Steve Grubb 已提交
113
#include <linux/audit.h>
114
#include <linux/dmaengine.h>
115
#include <linux/err.h>
116
#include <linux/ctype.h>
117
#include <linux/if_arp.h>
118
#include <linux/if_vlan.h>
119
#include <linux/ip.h>
120
#include <net/ip.h>
121 122
#include <linux/ipv6.h>
#include <linux/in.h>
D
David S. Miller 已提交
123 124
#include <linux/jhash.h>
#include <linux/random.h>
125
#include <trace/events/napi.h>
126
#include <trace/events/net.h>
127
#include <trace/events/skb.h>
128
#include <linux/pci.h>
129
#include <linux/inetdevice.h>
130
#include <linux/cpu_rmap.h>
131
#include <linux/static_key.h>
L
Linus Torvalds 已提交
132

133 134
#include "net-sysfs.h"

135 136 137
/* Instead of increasing this, you should create a hash table. */
#define MAX_GRO_SKBS 8

H
Herbert Xu 已提交
138 139 140
/* This should be increased if a protocol with a bigger head is added. */
#define GRO_MAX_HEAD (MAX_HEADER + 128)

L
Linus Torvalds 已提交
141
static DEFINE_SPINLOCK(ptype_lock);
142
static DEFINE_SPINLOCK(offload_lock);
143 144
struct list_head ptype_base[PTYPE_HASH_SIZE] __read_mostly;
struct list_head ptype_all __read_mostly;	/* Taps */
145
static struct list_head offload_base __read_mostly;
L
Linus Torvalds 已提交
146 147

/*
148
 * The @dev_base_head list is protected by @dev_base_lock and the rtnl
L
Linus Torvalds 已提交
149 150
 * semaphore.
 *
151
 * Pure readers hold dev_base_lock for reading, or rcu_read_lock()
L
Linus Torvalds 已提交
152 153
 *
 * Writers must hold the rtnl semaphore while they loop through the
154
 * dev_base_head list, and hold dev_base_lock for writing when they do the
L
Linus Torvalds 已提交
155 156 157 158 159 160 161 162 163 164 165 166 167 168
 * actual updates.  This allows pure readers to access the list even
 * while a writer is preparing to update it.
 *
 * To put it another way, dev_base_lock is held for writing only to
 * protect against pure readers; the rtnl semaphore provides the
 * protection against other writers.
 *
 * See, for example usages, register_netdevice() and
 * unregister_netdevice(), which must be called with the rtnl
 * semaphore held.
 */
DEFINE_RWLOCK(dev_base_lock);
EXPORT_SYMBOL(dev_base_lock);

169
seqcount_t devnet_rename_seq;
170

171 172 173 174 175
static inline void dev_base_seq_inc(struct net *net)
{
	while (++net->dev_base_seq == 0);
}

176
static inline struct hlist_head *dev_name_hash(struct net *net, const char *name)
L
Linus Torvalds 已提交
177
{
178 179
	unsigned int hash = full_name_hash(name, strnlen(name, IFNAMSIZ));

180
	return &net->dev_name_head[hash_32(hash, NETDEV_HASHBITS)];
L
Linus Torvalds 已提交
181 182
}

183
static inline struct hlist_head *dev_index_hash(struct net *net, int ifindex)
L
Linus Torvalds 已提交
184
{
185
	return &net->dev_index_head[ifindex & (NETDEV_HASHENTRIES - 1)];
L
Linus Torvalds 已提交
186 187
}

E
Eric Dumazet 已提交
188
static inline void rps_lock(struct softnet_data *sd)
189 190
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
191
	spin_lock(&sd->input_pkt_queue.lock);
192 193 194
#endif
}

E
Eric Dumazet 已提交
195
static inline void rps_unlock(struct softnet_data *sd)
196 197
{
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
198
	spin_unlock(&sd->input_pkt_queue.lock);
199 200 201
#endif
}

202 203 204
/* Device list insertion */
static int list_netdevice(struct net_device *dev)
{
205
	struct net *net = dev_net(dev);
206 207 208 209

	ASSERT_RTNL();

	write_lock_bh(&dev_base_lock);
210
	list_add_tail_rcu(&dev->dev_list, &net->dev_base_head);
211
	hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
212 213
	hlist_add_head_rcu(&dev->index_hlist,
			   dev_index_hash(net, dev->ifindex));
214
	write_unlock_bh(&dev_base_lock);
215 216 217

	dev_base_seq_inc(net);

218 219 220
	return 0;
}

221 222 223
/* Device list removal
 * caller must respect a RCU grace period before freeing/reusing dev
 */
224 225 226 227 228 229
static void unlist_netdevice(struct net_device *dev)
{
	ASSERT_RTNL();

	/* Unlink dev from the device chain */
	write_lock_bh(&dev_base_lock);
230
	list_del_rcu(&dev->dev_list);
231
	hlist_del_rcu(&dev->name_hlist);
232
	hlist_del_rcu(&dev->index_hlist);
233
	write_unlock_bh(&dev_base_lock);
234 235

	dev_base_seq_inc(dev_net(dev));
236 237
}

L
Linus Torvalds 已提交
238 239 240 241
/*
 *	Our notifier list
 */

242
static RAW_NOTIFIER_HEAD(netdev_chain);
L
Linus Torvalds 已提交
243 244 245 246 247

/*
 *	Device drivers call our routines to queue packets here. We empty the
 *	queue in the local softnet handler.
 */
248

249
DEFINE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
E
Eric Dumazet 已提交
250
EXPORT_PER_CPU_SYMBOL(softnet_data);
L
Linus Torvalds 已提交
251

252
#ifdef CONFIG_LOCKDEP
253
/*
254
 * register_netdevice() inits txq->_xmit_lock and sets lockdep class
255 256 257 258 259 260 261 262 263 264 265 266 267 268 269
 * according to dev->type
 */
static const unsigned short netdev_lock_type[] =
	{ARPHRD_NETROM, ARPHRD_ETHER, ARPHRD_EETHER, ARPHRD_AX25,
	 ARPHRD_PRONET, ARPHRD_CHAOS, ARPHRD_IEEE802, ARPHRD_ARCNET,
	 ARPHRD_APPLETLK, ARPHRD_DLCI, ARPHRD_ATM, ARPHRD_METRICOM,
	 ARPHRD_IEEE1394, ARPHRD_EUI64, ARPHRD_INFINIBAND, ARPHRD_SLIP,
	 ARPHRD_CSLIP, ARPHRD_SLIP6, ARPHRD_CSLIP6, ARPHRD_RSRVD,
	 ARPHRD_ADAPT, ARPHRD_ROSE, ARPHRD_X25, ARPHRD_HWX25,
	 ARPHRD_PPP, ARPHRD_CISCO, ARPHRD_LAPB, ARPHRD_DDCMP,
	 ARPHRD_RAWHDLC, ARPHRD_TUNNEL, ARPHRD_TUNNEL6, ARPHRD_FRAD,
	 ARPHRD_SKIP, ARPHRD_LOOPBACK, ARPHRD_LOCALTLK, ARPHRD_FDDI,
	 ARPHRD_BIF, ARPHRD_SIT, ARPHRD_IPDDP, ARPHRD_IPGRE,
	 ARPHRD_PIMREG, ARPHRD_HIPPI, ARPHRD_ASH, ARPHRD_ECONET,
	 ARPHRD_IRDA, ARPHRD_FCPP, ARPHRD_FCAL, ARPHRD_FCPL,
270 271 272
	 ARPHRD_FCFABRIC, ARPHRD_IEEE80211, ARPHRD_IEEE80211_PRISM,
	 ARPHRD_IEEE80211_RADIOTAP, ARPHRD_PHONET, ARPHRD_PHONET_PIPE,
	 ARPHRD_IEEE802154, ARPHRD_VOID, ARPHRD_NONE};
273

274
static const char *const netdev_lock_name[] =
275 276 277 278 279 280 281 282 283 284 285 286
	{"_xmit_NETROM", "_xmit_ETHER", "_xmit_EETHER", "_xmit_AX25",
	 "_xmit_PRONET", "_xmit_CHAOS", "_xmit_IEEE802", "_xmit_ARCNET",
	 "_xmit_APPLETLK", "_xmit_DLCI", "_xmit_ATM", "_xmit_METRICOM",
	 "_xmit_IEEE1394", "_xmit_EUI64", "_xmit_INFINIBAND", "_xmit_SLIP",
	 "_xmit_CSLIP", "_xmit_SLIP6", "_xmit_CSLIP6", "_xmit_RSRVD",
	 "_xmit_ADAPT", "_xmit_ROSE", "_xmit_X25", "_xmit_HWX25",
	 "_xmit_PPP", "_xmit_CISCO", "_xmit_LAPB", "_xmit_DDCMP",
	 "_xmit_RAWHDLC", "_xmit_TUNNEL", "_xmit_TUNNEL6", "_xmit_FRAD",
	 "_xmit_SKIP", "_xmit_LOOPBACK", "_xmit_LOCALTLK", "_xmit_FDDI",
	 "_xmit_BIF", "_xmit_SIT", "_xmit_IPDDP", "_xmit_IPGRE",
	 "_xmit_PIMREG", "_xmit_HIPPI", "_xmit_ASH", "_xmit_ECONET",
	 "_xmit_IRDA", "_xmit_FCPP", "_xmit_FCAL", "_xmit_FCPL",
287 288 289
	 "_xmit_FCFABRIC", "_xmit_IEEE80211", "_xmit_IEEE80211_PRISM",
	 "_xmit_IEEE80211_RADIOTAP", "_xmit_PHONET", "_xmit_PHONET_PIPE",
	 "_xmit_IEEE802154", "_xmit_VOID", "_xmit_NONE"};
290 291

static struct lock_class_key netdev_xmit_lock_key[ARRAY_SIZE(netdev_lock_type)];
292
static struct lock_class_key netdev_addr_lock_key[ARRAY_SIZE(netdev_lock_type)];
293 294 295 296 297 298 299 300 301 302 303 304

static inline unsigned short netdev_lock_pos(unsigned short dev_type)
{
	int i;

	for (i = 0; i < ARRAY_SIZE(netdev_lock_type); i++)
		if (netdev_lock_type[i] == dev_type)
			return i;
	/* the last key is used by default */
	return ARRAY_SIZE(netdev_lock_type) - 1;
}

305 306
static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
						 unsigned short dev_type)
307 308 309 310 311 312 313
{
	int i;

	i = netdev_lock_pos(dev_type);
	lockdep_set_class_and_name(lock, &netdev_xmit_lock_key[i],
				   netdev_lock_name[i]);
}
314 315 316 317 318 319 320 321 322 323

static inline void netdev_set_addr_lockdep_class(struct net_device *dev)
{
	int i;

	i = netdev_lock_pos(dev->type);
	lockdep_set_class_and_name(&dev->addr_list_lock,
				   &netdev_addr_lock_key[i],
				   netdev_lock_name[i]);
}
324
#else
325 326 327 328 329
static inline void netdev_set_xmit_lockdep_class(spinlock_t *lock,
						 unsigned short dev_type)
{
}
static inline void netdev_set_addr_lockdep_class(struct net_device *dev)
330 331 332
{
}
#endif
L
Linus Torvalds 已提交
333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355

/*******************************************************************************

		Protocol management and registration routines

*******************************************************************************/

/*
 *	Add a protocol ID to the list. Now that the input handler is
 *	smarter we can dispense with all the messy stuff that used to be
 *	here.
 *
 *	BEWARE!!! Protocol handlers, mangling input packets,
 *	MUST BE last in hash buckets and checking protocol handlers
 *	MUST start from promiscuous ptype_all chain in net_bh.
 *	It is true now, do not change it.
 *	Explanation follows: if protocol handler, mangling packet, will
 *	be the first on list, it is not able to sense, that packet
 *	is cloned and should be copied-on-write, so that it will
 *	change it and subsequent readers will get broken packet.
 *							--ANK (980803)
 */

356 357 358 359 360 361 362 363
static inline struct list_head *ptype_head(const struct packet_type *pt)
{
	if (pt->type == htons(ETH_P_ALL))
		return &ptype_all;
	else
		return &ptype_base[ntohs(pt->type) & PTYPE_HASH_MASK];
}

L
Linus Torvalds 已提交
364 365 366 367 368 369 370 371
/**
 *	dev_add_pack - add packet handler
 *	@pt: packet type declaration
 *
 *	Add a protocol handler to the networking stack. The passed &packet_type
 *	is linked into kernel lists and may not be freed until it has been
 *	removed from the kernel lists.
 *
372
 *	This call does not sleep therefore it can not
L
Linus Torvalds 已提交
373 374 375 376 377 378
 *	guarantee all CPU's that are in middle of receiving packets
 *	will see the new packet type (until the next received packet).
 */

void dev_add_pack(struct packet_type *pt)
{
379
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
380

381 382 383
	spin_lock(&ptype_lock);
	list_add_rcu(&pt->list, head);
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
384
}
E
Eric Dumazet 已提交
385
EXPORT_SYMBOL(dev_add_pack);
L
Linus Torvalds 已提交
386 387 388 389 390 391 392 393

/**
 *	__dev_remove_pack	 - remove packet handler
 *	@pt: packet type declaration
 *
 *	Remove a protocol handler that was previously added to the kernel
 *	protocol handlers by dev_add_pack(). The passed &packet_type is removed
 *	from the kernel lists and can be freed or reused once this function
394
 *	returns.
L
Linus Torvalds 已提交
395 396 397 398 399 400 401
 *
 *      The packet type might still be in use by receivers
 *	and must not be freed until after all the CPU's have gone
 *	through a quiescent state.
 */
void __dev_remove_pack(struct packet_type *pt)
{
402
	struct list_head *head = ptype_head(pt);
L
Linus Torvalds 已提交
403 404
	struct packet_type *pt1;

405
	spin_lock(&ptype_lock);
L
Linus Torvalds 已提交
406 407 408 409 410 411 412 413

	list_for_each_entry(pt1, head, list) {
		if (pt == pt1) {
			list_del_rcu(&pt->list);
			goto out;
		}
	}

414
	pr_warn("dev_remove_pack: %p not found\n", pt);
L
Linus Torvalds 已提交
415
out:
416
	spin_unlock(&ptype_lock);
L
Linus Torvalds 已提交
417
}
E
Eric Dumazet 已提交
418 419
EXPORT_SYMBOL(__dev_remove_pack);

L
Linus Torvalds 已提交
420 421 422 423 424 425 426 427 428 429 430 431 432 433 434
/**
 *	dev_remove_pack	 - remove packet handler
 *	@pt: packet type declaration
 *
 *	Remove a protocol handler that was previously added to the kernel
 *	protocol handlers by dev_add_pack(). The passed &packet_type is removed
 *	from the kernel lists and can be freed or reused once this function
 *	returns.
 *
 *	This call sleeps to guarantee that no CPU is looking at the packet
 *	type after return.
 */
void dev_remove_pack(struct packet_type *pt)
{
	__dev_remove_pack(pt);
435

L
Linus Torvalds 已提交
436 437
	synchronize_net();
}
E
Eric Dumazet 已提交
438
EXPORT_SYMBOL(dev_remove_pack);
L
Linus Torvalds 已提交
439

440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480

/**
 *	dev_add_offload - register offload handlers
 *	@po: protocol offload declaration
 *
 *	Add protocol offload handlers to the networking stack. The passed
 *	&proto_offload is linked into kernel lists and may not be freed until
 *	it has been removed from the kernel lists.
 *
 *	This call does not sleep therefore it can not
 *	guarantee all CPU's that are in middle of receiving packets
 *	will see the new offload handlers (until the next received packet).
 */
void dev_add_offload(struct packet_offload *po)
{
	struct list_head *head = &offload_base;

	spin_lock(&offload_lock);
	list_add_rcu(&po->list, head);
	spin_unlock(&offload_lock);
}
EXPORT_SYMBOL(dev_add_offload);

/**
 *	__dev_remove_offload	 - remove offload handler
 *	@po: packet offload declaration
 *
 *	Remove a protocol offload handler that was previously added to the
 *	kernel offload handlers by dev_add_offload(). The passed &offload_type
 *	is removed from the kernel lists and can be freed or reused once this
 *	function returns.
 *
 *      The packet type might still be in use by receivers
 *	and must not be freed until after all the CPU's have gone
 *	through a quiescent state.
 */
void __dev_remove_offload(struct packet_offload *po)
{
	struct list_head *head = &offload_base;
	struct packet_offload *po1;

481
	spin_lock(&offload_lock);
482 483 484 485 486 487 488 489 490 491

	list_for_each_entry(po1, head, list) {
		if (po == po1) {
			list_del_rcu(&po->list);
			goto out;
		}
	}

	pr_warn("dev_remove_offload: %p not found\n", po);
out:
492
	spin_unlock(&offload_lock);
493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515
}
EXPORT_SYMBOL(__dev_remove_offload);

/**
 *	dev_remove_offload	 - remove packet offload handler
 *	@po: packet offload declaration
 *
 *	Remove a packet offload handler that was previously added to the kernel
 *	offload handlers by dev_add_offload(). The passed &offload_type is
 *	removed from the kernel lists and can be freed or reused once this
 *	function returns.
 *
 *	This call sleeps to guarantee that no CPU is looking at the packet
 *	type after return.
 */
void dev_remove_offload(struct packet_offload *po)
{
	__dev_remove_offload(po);

	synchronize_net();
}
EXPORT_SYMBOL(dev_remove_offload);

L
Linus Torvalds 已提交
516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542
/******************************************************************************

		      Device Boot-time Settings Routines

*******************************************************************************/

/* Boot time configuration table */
static struct netdev_boot_setup dev_boot_setup[NETDEV_BOOT_SETUP_MAX];

/**
 *	netdev_boot_setup_add	- add new setup entry
 *	@name: name of the device
 *	@map: configured settings for the device
 *
 *	Adds new setup entry to the dev_boot_setup list.  The function
 *	returns 0 on error and 1 on success.  This is a generic routine to
 *	all netdevices.
 */
static int netdev_boot_setup_add(char *name, struct ifmap *map)
{
	struct netdev_boot_setup *s;
	int i;

	s = dev_boot_setup;
	for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) {
		if (s[i].name[0] == '\0' || s[i].name[0] == ' ') {
			memset(s[i].name, 0, sizeof(s[i].name));
543
			strlcpy(s[i].name, name, IFNAMSIZ);
L
Linus Torvalds 已提交
544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567
			memcpy(&s[i].map, map, sizeof(s[i].map));
			break;
		}
	}

	return i >= NETDEV_BOOT_SETUP_MAX ? 0 : 1;
}

/**
 *	netdev_boot_setup_check	- check boot time settings
 *	@dev: the netdevice
 *
 * 	Check boot time settings for the device.
 *	The found settings are set for the device to be used
 *	later in the device probing.
 *	Returns 0 if no settings found, 1 if they are.
 */
int netdev_boot_setup_check(struct net_device *dev)
{
	struct netdev_boot_setup *s = dev_boot_setup;
	int i;

	for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++) {
		if (s[i].name[0] != '\0' && s[i].name[0] != ' ' &&
568
		    !strcmp(dev->name, s[i].name)) {
L
Linus Torvalds 已提交
569 570 571 572 573 574 575 576 577
			dev->irq 	= s[i].map.irq;
			dev->base_addr 	= s[i].map.base_addr;
			dev->mem_start 	= s[i].map.mem_start;
			dev->mem_end 	= s[i].map.mem_end;
			return 1;
		}
	}
	return 0;
}
E
Eric Dumazet 已提交
578
EXPORT_SYMBOL(netdev_boot_setup_check);
L
Linus Torvalds 已提交
579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602


/**
 *	netdev_boot_base	- get address from boot time settings
 *	@prefix: prefix for network device
 *	@unit: id for network device
 *
 * 	Check boot time settings for the base address of device.
 *	The found settings are set for the device to be used
 *	later in the device probing.
 *	Returns 0 if no settings found.
 */
unsigned long netdev_boot_base(const char *prefix, int unit)
{
	const struct netdev_boot_setup *s = dev_boot_setup;
	char name[IFNAMSIZ];
	int i;

	sprintf(name, "%s%d", prefix, unit);

	/*
	 * If device already registered then return base of 1
	 * to indicate not to probe for this interface
	 */
603
	if (__dev_get_by_name(&init_net, name))
L
Linus Torvalds 已提交
604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648
		return 1;

	for (i = 0; i < NETDEV_BOOT_SETUP_MAX; i++)
		if (!strcmp(name, s[i].name))
			return s[i].map.base_addr;
	return 0;
}

/*
 * Saves at boot time configured settings for any netdevice.
 */
int __init netdev_boot_setup(char *str)
{
	int ints[5];
	struct ifmap map;

	str = get_options(str, ARRAY_SIZE(ints), ints);
	if (!str || !*str)
		return 0;

	/* Save settings */
	memset(&map, 0, sizeof(map));
	if (ints[0] > 0)
		map.irq = ints[1];
	if (ints[0] > 1)
		map.base_addr = ints[2];
	if (ints[0] > 2)
		map.mem_start = ints[3];
	if (ints[0] > 3)
		map.mem_end = ints[4];

	/* Add new entry to the list */
	return netdev_boot_setup_add(str, &map);
}

__setup("netdev=", netdev_boot_setup);

/*******************************************************************************

			    Device Interface Subroutines

*******************************************************************************/

/**
 *	__dev_get_by_name	- find a device by its name
649
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
650 651 652 653 654 655 656 657 658
 *	@name: name to find
 *
 *	Find an interface by name. Must be called under RTNL semaphore
 *	or @dev_base_lock. If the name is found a pointer to the device
 *	is returned. If the name is not found then %NULL is returned. The
 *	reference counters are not incremented so the caller must be
 *	careful with locks.
 */

659
struct net_device *__dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
660
{
E
Eric Dumazet 已提交
661 662
	struct net_device *dev;
	struct hlist_head *head = dev_name_hash(net, name);
L
Linus Torvalds 已提交
663

664
	hlist_for_each_entry(dev, head, name_hlist)
L
Linus Torvalds 已提交
665 666
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;
E
Eric Dumazet 已提交
667

L
Linus Torvalds 已提交
668 669
	return NULL;
}
E
Eric Dumazet 已提交
670
EXPORT_SYMBOL(__dev_get_by_name);
L
Linus Torvalds 已提交
671

672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688
/**
 *	dev_get_by_name_rcu	- find a device by its name
 *	@net: the applicable net namespace
 *	@name: name to find
 *
 *	Find an interface by name.
 *	If the name is found a pointer to the device is returned.
 * 	If the name is not found then %NULL is returned.
 *	The reference counters are not incremented so the caller must be
 *	careful with locks. The caller must hold RCU lock.
 */

struct net_device *dev_get_by_name_rcu(struct net *net, const char *name)
{
	struct net_device *dev;
	struct hlist_head *head = dev_name_hash(net, name);

689
	hlist_for_each_entry_rcu(dev, head, name_hlist)
690 691 692 693 694 695 696
		if (!strncmp(dev->name, name, IFNAMSIZ))
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_name_rcu);

L
Linus Torvalds 已提交
697 698
/**
 *	dev_get_by_name		- find a device by its name
699
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
700 701 702 703 704 705 706 707 708
 *	@name: name to find
 *
 *	Find an interface by name. This can be called from any
 *	context and does its own locking. The returned handle has
 *	the usage count incremented and the caller must use dev_put() to
 *	release it when it is no longer needed. %NULL is returned if no
 *	matching device is found.
 */

709
struct net_device *dev_get_by_name(struct net *net, const char *name)
L
Linus Torvalds 已提交
710 711 712
{
	struct net_device *dev;

713 714
	rcu_read_lock();
	dev = dev_get_by_name_rcu(net, name);
L
Linus Torvalds 已提交
715 716
	if (dev)
		dev_hold(dev);
717
	rcu_read_unlock();
L
Linus Torvalds 已提交
718 719
	return dev;
}
E
Eric Dumazet 已提交
720
EXPORT_SYMBOL(dev_get_by_name);
L
Linus Torvalds 已提交
721 722 723

/**
 *	__dev_get_by_index - find a device by its ifindex
724
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
725 726 727 728 729 730 731 732 733
 *	@ifindex: index of device
 *
 *	Search for an interface by index. Returns %NULL if the device
 *	is not found or a pointer to the device. The device has not
 *	had its reference counter increased so the caller must be careful
 *	about locking. The caller must hold either the RTNL semaphore
 *	or @dev_base_lock.
 */

734
struct net_device *__dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
735
{
E
Eric Dumazet 已提交
736 737
	struct net_device *dev;
	struct hlist_head *head = dev_index_hash(net, ifindex);
L
Linus Torvalds 已提交
738

739
	hlist_for_each_entry(dev, head, index_hlist)
L
Linus Torvalds 已提交
740 741
		if (dev->ifindex == ifindex)
			return dev;
E
Eric Dumazet 已提交
742

L
Linus Torvalds 已提交
743 744
	return NULL;
}
E
Eric Dumazet 已提交
745
EXPORT_SYMBOL(__dev_get_by_index);
L
Linus Torvalds 已提交
746

747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762
/**
 *	dev_get_by_index_rcu - find a device by its ifindex
 *	@net: the applicable net namespace
 *	@ifindex: index of device
 *
 *	Search for an interface by index. Returns %NULL if the device
 *	is not found or a pointer to the device. The device has not
 *	had its reference counter increased so the caller must be careful
 *	about locking. The caller must hold RCU lock.
 */

struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex)
{
	struct net_device *dev;
	struct hlist_head *head = dev_index_hash(net, ifindex);

763
	hlist_for_each_entry_rcu(dev, head, index_hlist)
764 765 766 767 768 769 770
		if (dev->ifindex == ifindex)
			return dev;

	return NULL;
}
EXPORT_SYMBOL(dev_get_by_index_rcu);

L
Linus Torvalds 已提交
771 772 773

/**
 *	dev_get_by_index - find a device by its ifindex
774
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
775 776 777 778 779 780 781 782
 *	@ifindex: index of device
 *
 *	Search for an interface by index. Returns NULL if the device
 *	is not found or a pointer to the device. The device returned has
 *	had a reference added and the pointer is safe until the user calls
 *	dev_put to indicate they have finished with it.
 */

783
struct net_device *dev_get_by_index(struct net *net, int ifindex)
L
Linus Torvalds 已提交
784 785 786
{
	struct net_device *dev;

787 788
	rcu_read_lock();
	dev = dev_get_by_index_rcu(net, ifindex);
L
Linus Torvalds 已提交
789 790
	if (dev)
		dev_hold(dev);
791
	rcu_read_unlock();
L
Linus Torvalds 已提交
792 793
	return dev;
}
E
Eric Dumazet 已提交
794
EXPORT_SYMBOL(dev_get_by_index);
L
Linus Torvalds 已提交
795 796

/**
797
 *	dev_getbyhwaddr_rcu - find a device by its hardware address
798
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
799 800 801 802
 *	@type: media type of device
 *	@ha: hardware address
 *
 *	Search for an interface by MAC address. Returns NULL if the device
E
Eric Dumazet 已提交
803 804
 *	is not found or a pointer to the device.
 *	The caller must hold RCU or RTNL.
805
 *	The returned device has not had its ref count increased
L
Linus Torvalds 已提交
806 807 808 809
 *	and the caller must therefore be careful about locking
 *
 */

810 811
struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
				       const char *ha)
L
Linus Torvalds 已提交
812 813 814
{
	struct net_device *dev;

815
	for_each_netdev_rcu(net, dev)
L
Linus Torvalds 已提交
816 817
		if (dev->type == type &&
		    !memcmp(dev->dev_addr, ha, dev->addr_len))
818 819 820
			return dev;

	return NULL;
L
Linus Torvalds 已提交
821
}
822
EXPORT_SYMBOL(dev_getbyhwaddr_rcu);
823

824
struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type)
L
Linus Torvalds 已提交
825 826 827
{
	struct net_device *dev;

828
	ASSERT_RTNL();
829
	for_each_netdev(net, dev)
830
		if (dev->type == type)
831 832 833
			return dev;

	return NULL;
834 835 836
}
EXPORT_SYMBOL(__dev_getfirstbyhwtype);

837
struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type)
838
{
839
	struct net_device *dev, *ret = NULL;
840

841 842 843 844 845 846 847 848 849
	rcu_read_lock();
	for_each_netdev_rcu(net, dev)
		if (dev->type == type) {
			dev_hold(dev);
			ret = dev;
			break;
		}
	rcu_read_unlock();
	return ret;
L
Linus Torvalds 已提交
850 851 852 853
}
EXPORT_SYMBOL(dev_getfirstbyhwtype);

/**
E
Eric Dumazet 已提交
854
 *	dev_get_by_flags_rcu - find any device with given flags
855
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
856 857 858 859
 *	@if_flags: IFF_* values
 *	@mask: bitmask of bits in if_flags to check
 *
 *	Search for any interface with the given flags. Returns NULL if a device
E
Eric Dumazet 已提交
860 861
 *	is not found or a pointer to the device. Must be called inside
 *	rcu_read_lock(), and result refcount is unchanged.
L
Linus Torvalds 已提交
862 863
 */

E
Eric Dumazet 已提交
864
struct net_device *dev_get_by_flags_rcu(struct net *net, unsigned short if_flags,
E
Eric Dumazet 已提交
865
				    unsigned short mask)
L
Linus Torvalds 已提交
866
{
867
	struct net_device *dev, *ret;
L
Linus Torvalds 已提交
868

869
	ret = NULL;
870
	for_each_netdev_rcu(net, dev) {
L
Linus Torvalds 已提交
871
		if (((dev->flags ^ if_flags) & mask) == 0) {
872
			ret = dev;
L
Linus Torvalds 已提交
873 874 875
			break;
		}
	}
876
	return ret;
L
Linus Torvalds 已提交
877
}
E
Eric Dumazet 已提交
878
EXPORT_SYMBOL(dev_get_by_flags_rcu);
L
Linus Torvalds 已提交
879 880 881 882 883 884

/**
 *	dev_valid_name - check if name is okay for network device
 *	@name: name string
 *
 *	Network device names need to be valid file names to
885 886
 *	to allow sysfs to work.  We also disallow any kind of
 *	whitespace.
L
Linus Torvalds 已提交
887
 */
888
bool dev_valid_name(const char *name)
L
Linus Torvalds 已提交
889
{
890
	if (*name == '\0')
891
		return false;
892
	if (strlen(name) >= IFNAMSIZ)
893
		return false;
894
	if (!strcmp(name, ".") || !strcmp(name, ".."))
895
		return false;
896 897 898

	while (*name) {
		if (*name == '/' || isspace(*name))
899
			return false;
900 901
		name++;
	}
902
	return true;
L
Linus Torvalds 已提交
903
}
E
Eric Dumazet 已提交
904
EXPORT_SYMBOL(dev_valid_name);
L
Linus Torvalds 已提交
905 906

/**
907 908
 *	__dev_alloc_name - allocate a name for a device
 *	@net: network namespace to allocate the device name in
L
Linus Torvalds 已提交
909
 *	@name: name format string
910
 *	@buf:  scratch buffer and result name string
L
Linus Torvalds 已提交
911 912
 *
 *	Passed a format string - eg "lt%d" it will try and find a suitable
S
Stephen Hemminger 已提交
913 914 915 916 917 918
 *	id. It scans list of devices to build up a free map, then chooses
 *	the first empty slot. The caller must hold the dev_base or rtnl lock
 *	while allocating the name and adding the device in order to avoid
 *	duplicates.
 *	Limited to bits_per_byte * page size devices (ie 32K on most platforms).
 *	Returns the number of the unit assigned or a negative errno code.
L
Linus Torvalds 已提交
919 920
 */

921
static int __dev_alloc_name(struct net *net, const char *name, char *buf)
L
Linus Torvalds 已提交
922 923 924 925
{
	int i = 0;
	const char *p;
	const int max_netdevices = 8*PAGE_SIZE;
S
Stephen Hemminger 已提交
926
	unsigned long *inuse;
L
Linus Torvalds 已提交
927 928 929 930 931 932 933 934 935 936 937 938 939
	struct net_device *d;

	p = strnchr(name, IFNAMSIZ-1, '%');
	if (p) {
		/*
		 * Verify the string as this thing may have come from
		 * the user.  There must be either one "%d" and no other "%"
		 * characters.
		 */
		if (p[1] != 'd' || strchr(p + 2, '%'))
			return -EINVAL;

		/* Use one page as a bit array of possible slots */
S
Stephen Hemminger 已提交
940
		inuse = (unsigned long *) get_zeroed_page(GFP_ATOMIC);
L
Linus Torvalds 已提交
941 942 943
		if (!inuse)
			return -ENOMEM;

944
		for_each_netdev(net, d) {
L
Linus Torvalds 已提交
945 946 947 948 949 950
			if (!sscanf(d->name, name, &i))
				continue;
			if (i < 0 || i >= max_netdevices)
				continue;

			/*  avoid cases where sscanf is not exact inverse of printf */
951
			snprintf(buf, IFNAMSIZ, name, i);
L
Linus Torvalds 已提交
952 953 954 955 956 957 958 959
			if (!strncmp(buf, d->name, IFNAMSIZ))
				set_bit(i, inuse);
		}

		i = find_first_zero_bit(inuse, max_netdevices);
		free_page((unsigned long) inuse);
	}

960 961
	if (buf != name)
		snprintf(buf, IFNAMSIZ, name, i);
962
	if (!__dev_get_by_name(net, buf))
L
Linus Torvalds 已提交
963 964 965 966 967 968 969 970 971
		return i;

	/* It is possible to run out of possible slots
	 * when the name is long and there isn't enough space left
	 * for the digits, or if all bits are used.
	 */
	return -ENFILE;
}

972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991
/**
 *	dev_alloc_name - allocate a name for a device
 *	@dev: device
 *	@name: name format string
 *
 *	Passed a format string - eg "lt%d" it will try and find a suitable
 *	id. It scans list of devices to build up a free map, then chooses
 *	the first empty slot. The caller must hold the dev_base or rtnl lock
 *	while allocating the name and adding the device in order to avoid
 *	duplicates.
 *	Limited to bits_per_byte * page size devices (ie 32K on most platforms).
 *	Returns the number of the unit assigned or a negative errno code.
 */

int dev_alloc_name(struct net_device *dev, const char *name)
{
	char buf[IFNAMSIZ];
	struct net *net;
	int ret;

992 993
	BUG_ON(!dev_net(dev));
	net = dev_net(dev);
994 995 996 997 998
	ret = __dev_alloc_name(net, name, buf);
	if (ret >= 0)
		strlcpy(dev->name, buf, IFNAMSIZ);
	return ret;
}
E
Eric Dumazet 已提交
999
EXPORT_SYMBOL(dev_alloc_name);
1000

1001 1002 1003
static int dev_alloc_name_ns(struct net *net,
			     struct net_device *dev,
			     const char *name)
1004
{
1005 1006
	char buf[IFNAMSIZ];
	int ret;
1007

1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018
	ret = __dev_alloc_name(net, name, buf);
	if (ret >= 0)
		strlcpy(dev->name, buf, IFNAMSIZ);
	return ret;
}

static int dev_get_valid_name(struct net *net,
			      struct net_device *dev,
			      const char *name)
{
	BUG_ON(!net);
1019

1020 1021 1022
	if (!dev_valid_name(name))
		return -EINVAL;

1023
	if (strchr(name, '%'))
1024
		return dev_alloc_name_ns(net, dev, name);
1025 1026
	else if (__dev_get_by_name(net, name))
		return -EEXIST;
1027 1028
	else if (dev->name != name)
		strlcpy(dev->name, name, IFNAMSIZ);
1029 1030 1031

	return 0;
}
L
Linus Torvalds 已提交
1032 1033 1034 1035 1036 1037 1038 1039 1040

/**
 *	dev_change_name - change name of a device
 *	@dev: device
 *	@newname: name (or format string) must be at least IFNAMSIZ
 *
 *	Change name of a device, can pass format strings "eth%d".
 *	for wildcarding.
 */
1041
int dev_change_name(struct net_device *dev, const char *newname)
L
Linus Torvalds 已提交
1042
{
1043
	char oldname[IFNAMSIZ];
L
Linus Torvalds 已提交
1044
	int err = 0;
1045
	int ret;
1046
	struct net *net;
L
Linus Torvalds 已提交
1047 1048

	ASSERT_RTNL();
1049
	BUG_ON(!dev_net(dev));
L
Linus Torvalds 已提交
1050

1051
	net = dev_net(dev);
L
Linus Torvalds 已提交
1052 1053 1054
	if (dev->flags & IFF_UP)
		return -EBUSY;

1055
	write_seqcount_begin(&devnet_rename_seq);
1056 1057

	if (strncmp(newname, dev->name, IFNAMSIZ) == 0) {
1058
		write_seqcount_end(&devnet_rename_seq);
1059
		return 0;
1060
	}
1061

1062 1063
	memcpy(oldname, dev->name, IFNAMSIZ);

1064
	err = dev_get_valid_name(net, dev, newname);
1065
	if (err < 0) {
1066
		write_seqcount_end(&devnet_rename_seq);
1067
		return err;
1068
	}
L
Linus Torvalds 已提交
1069

1070
rollback:
1071 1072 1073
	ret = device_rename(&dev->dev, dev->name);
	if (ret) {
		memcpy(dev->name, oldname, IFNAMSIZ);
1074
		write_seqcount_end(&devnet_rename_seq);
1075
		return ret;
1076
	}
1077

1078
	write_seqcount_end(&devnet_rename_seq);
1079

1080
	write_lock_bh(&dev_base_lock);
1081
	hlist_del_rcu(&dev->name_hlist);
1082 1083 1084 1085 1086 1087
	write_unlock_bh(&dev_base_lock);

	synchronize_rcu();

	write_lock_bh(&dev_base_lock);
	hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
1088 1089
	write_unlock_bh(&dev_base_lock);

1090
	ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev);
1091 1092 1093
	ret = notifier_to_errno(ret);

	if (ret) {
1094 1095
		/* err >= 0 after dev_alloc_name() or stores the first errno */
		if (err >= 0) {
1096
			err = ret;
1097
			write_seqcount_begin(&devnet_rename_seq);
1098 1099
			memcpy(dev->name, oldname, IFNAMSIZ);
			goto rollback;
1100
		} else {
1101
			pr_err("%s: name change rollback failed: %d\n",
1102
			       dev->name, ret);
1103 1104
		}
	}
L
Linus Torvalds 已提交
1105 1106 1107 1108

	return err;
}

1109 1110 1111 1112
/**
 *	dev_set_alias - change ifalias of a device
 *	@dev: device
 *	@alias: name up to IFALIASZ
1113
 *	@len: limit of bytes to copy from info
1114 1115 1116 1117 1118
 *
 *	Set ifalias for a device,
 */
int dev_set_alias(struct net_device *dev, const char *alias, size_t len)
{
1119 1120
	char *new_ifalias;

1121 1122 1123 1124 1125
	ASSERT_RTNL();

	if (len >= IFALIASZ)
		return -EINVAL;

1126
	if (!len) {
1127 1128
		kfree(dev->ifalias);
		dev->ifalias = NULL;
1129 1130 1131
		return 0;
	}

1132 1133
	new_ifalias = krealloc(dev->ifalias, len + 1, GFP_KERNEL);
	if (!new_ifalias)
1134
		return -ENOMEM;
1135
	dev->ifalias = new_ifalias;
1136 1137 1138 1139 1140 1141

	strlcpy(dev->ifalias, alias, len+1);
	return len;
}


1142
/**
S
Stephen Hemminger 已提交
1143
 *	netdev_features_change - device changes features
1144 1145 1146 1147 1148 1149
 *	@dev: device to cause notification
 *
 *	Called to indicate a device has changed features.
 */
void netdev_features_change(struct net_device *dev)
{
1150
	call_netdevice_notifiers(NETDEV_FEAT_CHANGE, dev);
1151 1152 1153
}
EXPORT_SYMBOL(netdev_features_change);

L
Linus Torvalds 已提交
1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164
/**
 *	netdev_state_change - device changes state
 *	@dev: device to cause notification
 *
 *	Called to indicate a device has changed state. This function calls
 *	the notifier chains for netdev_chain and sends a NEWLINK message
 *	to the routing socket.
 */
void netdev_state_change(struct net_device *dev)
{
	if (dev->flags & IFF_UP) {
1165
		call_netdevice_notifiers(NETDEV_CHANGE, dev);
L
Linus Torvalds 已提交
1166 1167 1168
		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
	}
}
E
Eric Dumazet 已提交
1169
EXPORT_SYMBOL(netdev_state_change);
L
Linus Torvalds 已提交
1170

1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181
/**
 * 	netdev_notify_peers - notify network peers about existence of @dev
 * 	@dev: network device
 *
 * Generate traffic such that interested network peers are aware of
 * @dev, such as by generating a gratuitous ARP. This may be used when
 * a device wants to inform the rest of the network about some sort of
 * reconfiguration such as a failover event or virtual machine
 * migration.
 */
void netdev_notify_peers(struct net_device *dev)
1182
{
1183 1184 1185
	rtnl_lock();
	call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, dev);
	rtnl_unlock();
1186
}
1187
EXPORT_SYMBOL(netdev_notify_peers);
1188

1189
static int __dev_open(struct net_device *dev)
L
Linus Torvalds 已提交
1190
{
1191
	const struct net_device_ops *ops = dev->netdev_ops;
1192
	int ret;
L
Linus Torvalds 已提交
1193

1194 1195
	ASSERT_RTNL();

L
Linus Torvalds 已提交
1196 1197 1198
	if (!netif_device_present(dev))
		return -ENODEV;

1199 1200 1201 1202 1203 1204 1205 1206
	/* Block netpoll from trying to do any rx path servicing.
	 * If we don't do this there is a chance ndo_poll_controller
	 * or ndo_poll may be running while we open the device
	 */
	ret = netpoll_rx_disable(dev);
	if (ret)
		return ret;

1207 1208 1209 1210 1211
	ret = call_netdevice_notifiers(NETDEV_PRE_UP, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		return ret;

L
Linus Torvalds 已提交
1212
	set_bit(__LINK_STATE_START, &dev->state);
1213

1214 1215
	if (ops->ndo_validate_addr)
		ret = ops->ndo_validate_addr(dev);
1216

1217 1218
	if (!ret && ops->ndo_open)
		ret = ops->ndo_open(dev);
L
Linus Torvalds 已提交
1219

1220 1221
	netpoll_rx_enable(dev);

1222 1223 1224
	if (ret)
		clear_bit(__LINK_STATE_START, &dev->state);
	else {
L
Linus Torvalds 已提交
1225
		dev->flags |= IFF_UP;
1226
		net_dmaengine_get();
1227
		dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
1228
		dev_activate(dev);
1229
		add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
1230
	}
1231

L
Linus Torvalds 已提交
1232 1233 1234 1235
	return ret;
}

/**
1236 1237
 *	dev_open	- prepare an interface for use.
 *	@dev:	device to open
L
Linus Torvalds 已提交
1238
 *
1239 1240 1241 1242 1243 1244 1245
 *	Takes a device from down to up state. The device's private open
 *	function is invoked and then the multicast lists are loaded. Finally
 *	the device is moved into the up state and a %NETDEV_UP message is
 *	sent to the netdev notifier chain.
 *
 *	Calling this function on an active interface is a nop. On a failure
 *	a negative errno code is returned.
L
Linus Torvalds 已提交
1246
 */
1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264
int dev_open(struct net_device *dev)
{
	int ret;

	if (dev->flags & IFF_UP)
		return 0;

	ret = __dev_open(dev);
	if (ret < 0)
		return ret;

	rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
	call_netdevice_notifiers(NETDEV_UP, dev);

	return ret;
}
EXPORT_SYMBOL(dev_open);

1265
static int __dev_close_many(struct list_head *head)
L
Linus Torvalds 已提交
1266
{
1267
	struct net_device *dev;
1268

1269
	ASSERT_RTNL();
1270 1271
	might_sleep();

1272 1273
	list_for_each_entry(dev, head, unreg_list) {
		call_netdevice_notifiers(NETDEV_GOING_DOWN, dev);
L
Linus Torvalds 已提交
1274

1275
		clear_bit(__LINK_STATE_START, &dev->state);
L
Linus Torvalds 已提交
1276

1277 1278 1279 1280 1281 1282 1283 1284
		/* Synchronize to scheduled poll. We cannot touch poll list, it
		 * can be even on different cpu. So just clear netif_running().
		 *
		 * dev->stop() will invoke napi_disable() on all of it's
		 * napi_struct instances on this device.
		 */
		smp_mb__after_clear_bit(); /* Commit netif_running(). */
	}
L
Linus Torvalds 已提交
1285

1286
	dev_deactivate_many(head);
1287

1288 1289
	list_for_each_entry(dev, head, unreg_list) {
		const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
1290

1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309
		/*
		 *	Call the device specific close. This cannot fail.
		 *	Only if device is UP
		 *
		 *	We allow it to be called even after a DETACH hot-plug
		 *	event.
		 */
		if (ops->ndo_stop)
			ops->ndo_stop(dev);

		dev->flags &= ~IFF_UP;
		net_dmaengine_put();
	}

	return 0;
}

static int __dev_close(struct net_device *dev)
{
1310
	int retval;
1311 1312
	LIST_HEAD(single);

1313 1314 1315 1316 1317
	/* Temporarily disable netpoll until the interface is down */
	retval = netpoll_rx_disable(dev);
	if (retval)
		return retval;

1318
	list_add(&dev->unreg_list, &single);
1319 1320
	retval = __dev_close_many(&single);
	list_del(&single);
1321 1322

	netpoll_rx_enable(dev);
1323
	return retval;
1324 1325
}

E
Eric Dumazet 已提交
1326
static int dev_close_many(struct list_head *head)
1327 1328 1329
{
	struct net_device *dev, *tmp;
	LIST_HEAD(tmp_list);
L
Linus Torvalds 已提交
1330

1331 1332 1333 1334 1335
	list_for_each_entry_safe(dev, tmp, head, unreg_list)
		if (!(dev->flags & IFF_UP))
			list_move(&dev->unreg_list, &tmp_list);

	__dev_close_many(head);
L
Linus Torvalds 已提交
1336

1337 1338 1339 1340
	list_for_each_entry(dev, head, unreg_list) {
		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
		call_netdevice_notifiers(NETDEV_DOWN, dev);
	}
1341

1342 1343
	/* rollback_registered_many needs the complete original list */
	list_splice(&tmp_list, head);
1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357
	return 0;
}

/**
 *	dev_close - shutdown an interface.
 *	@dev: device to shutdown
 *
 *	This function moves an active device into down state. A
 *	%NETDEV_GOING_DOWN is sent to the netdev notifier chain. The device
 *	is then deactivated and finally a %NETDEV_DOWN is sent to the notifier
 *	chain.
 */
int dev_close(struct net_device *dev)
{
1358
	int ret = 0;
1359 1360
	if (dev->flags & IFF_UP) {
		LIST_HEAD(single);
L
Linus Torvalds 已提交
1361

1362 1363 1364 1365 1366
		/* Block netpoll rx while the interface is going down */
		ret = netpoll_rx_disable(dev);
		if (ret)
			return ret;

1367 1368 1369
		list_add(&dev->unreg_list, &single);
		dev_close_many(&single);
		list_del(&single);
1370 1371

		netpoll_rx_enable(dev);
1372
	}
1373
	return ret;
L
Linus Torvalds 已提交
1374
}
E
Eric Dumazet 已提交
1375
EXPORT_SYMBOL(dev_close);
L
Linus Torvalds 已提交
1376 1377


1378 1379 1380 1381 1382 1383 1384 1385 1386 1387
/**
 *	dev_disable_lro - disable Large Receive Offload on a device
 *	@dev: device
 *
 *	Disable Large Receive Offload (LRO) on a net device.  Must be
 *	called under RTNL.  This is needed if received packets may be
 *	forwarded to another interface.
 */
void dev_disable_lro(struct net_device *dev)
{
1388 1389 1390 1391 1392 1393 1394
	/*
	 * If we're trying to disable lro on a vlan device
	 * use the underlying physical device instead
	 */
	if (is_vlan_dev(dev))
		dev = vlan_dev_real_dev(dev);

M
Michał Mirosław 已提交
1395 1396
	dev->wanted_features &= ~NETIF_F_LRO;
	netdev_update_features(dev);
1397

1398 1399
	if (unlikely(dev->features & NETIF_F_LRO))
		netdev_WARN(dev, "failed to disable LRO!\n");
1400 1401 1402 1403
}
EXPORT_SYMBOL(dev_disable_lro);


1404 1405
static int dev_boot_phase = 1;

L
Linus Torvalds 已提交
1406 1407 1408 1409 1410 1411 1412 1413 1414 1415
/**
 *	register_netdevice_notifier - register a network notifier block
 *	@nb: notifier
 *
 *	Register a notifier to be called when network device events occur.
 *	The notifier passed is linked into the kernel structures and must
 *	not be reused until it has been unregistered. A negative errno code
 *	is returned on a failure.
 *
 * 	When registered all registration and up events are replayed
1416
 *	to the new notifier to allow device to have a race free
L
Linus Torvalds 已提交
1417 1418 1419 1420 1421 1422
 *	view of the network device list.
 */

int register_netdevice_notifier(struct notifier_block *nb)
{
	struct net_device *dev;
1423
	struct net_device *last;
1424
	struct net *net;
L
Linus Torvalds 已提交
1425 1426 1427
	int err;

	rtnl_lock();
1428
	err = raw_notifier_chain_register(&netdev_chain, nb);
1429 1430
	if (err)
		goto unlock;
1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441
	if (dev_boot_phase)
		goto unlock;
	for_each_net(net) {
		for_each_netdev(net, dev) {
			err = nb->notifier_call(nb, NETDEV_REGISTER, dev);
			err = notifier_to_errno(err);
			if (err)
				goto rollback;

			if (!(dev->flags & IFF_UP))
				continue;
L
Linus Torvalds 已提交
1442

1443 1444
			nb->notifier_call(nb, NETDEV_UP, dev);
		}
L
Linus Torvalds 已提交
1445
	}
1446 1447

unlock:
L
Linus Torvalds 已提交
1448 1449
	rtnl_unlock();
	return err;
1450 1451 1452

rollback:
	last = dev;
1453 1454 1455
	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev == last)
1456
				goto outroll;
1457

1458 1459 1460 1461 1462
			if (dev->flags & IFF_UP) {
				nb->notifier_call(nb, NETDEV_GOING_DOWN, dev);
				nb->notifier_call(nb, NETDEV_DOWN, dev);
			}
			nb->notifier_call(nb, NETDEV_UNREGISTER, dev);
1463 1464
		}
	}
1465

1466
outroll:
1467
	raw_notifier_chain_unregister(&netdev_chain, nb);
1468
	goto unlock;
L
Linus Torvalds 已提交
1469
}
E
Eric Dumazet 已提交
1470
EXPORT_SYMBOL(register_netdevice_notifier);
L
Linus Torvalds 已提交
1471 1472 1473 1474 1475 1476 1477 1478 1479

/**
 *	unregister_netdevice_notifier - unregister a network notifier block
 *	@nb: notifier
 *
 *	Unregister a notifier previously registered by
 *	register_netdevice_notifier(). The notifier is unlinked into the
 *	kernel structures and may then be reused. A negative errno code
 *	is returned on a failure.
1480 1481 1482 1483
 *
 * 	After unregistering unregister and down device events are synthesized
 *	for all devices on the device list to the removed notifier to remove
 *	the need for special case cleanup code.
L
Linus Torvalds 已提交
1484 1485 1486 1487
 */

int unregister_netdevice_notifier(struct notifier_block *nb)
{
1488 1489
	struct net_device *dev;
	struct net *net;
1490 1491 1492
	int err;

	rtnl_lock();
1493
	err = raw_notifier_chain_unregister(&netdev_chain, nb);
1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506
	if (err)
		goto unlock;

	for_each_net(net) {
		for_each_netdev(net, dev) {
			if (dev->flags & IFF_UP) {
				nb->notifier_call(nb, NETDEV_GOING_DOWN, dev);
				nb->notifier_call(nb, NETDEV_DOWN, dev);
			}
			nb->notifier_call(nb, NETDEV_UNREGISTER, dev);
		}
	}
unlock:
1507 1508
	rtnl_unlock();
	return err;
L
Linus Torvalds 已提交
1509
}
E
Eric Dumazet 已提交
1510
EXPORT_SYMBOL(unregister_netdevice_notifier);
L
Linus Torvalds 已提交
1511 1512 1513 1514

/**
 *	call_netdevice_notifiers - call all network notifier blocks
 *      @val: value passed unmodified to notifier function
1515
 *      @dev: net_device pointer passed unmodified to notifier function
L
Linus Torvalds 已提交
1516 1517
 *
 *	Call all network notifier blocks.  Parameters and return value
1518
 *	are as for raw_notifier_call_chain().
L
Linus Torvalds 已提交
1519 1520
 */

1521
int call_netdevice_notifiers(unsigned long val, struct net_device *dev)
L
Linus Torvalds 已提交
1522
{
1523
	ASSERT_RTNL();
1524
	return raw_notifier_call_chain(&netdev_chain, val, dev);
L
Linus Torvalds 已提交
1525
}
1526
EXPORT_SYMBOL(call_netdevice_notifiers);
L
Linus Torvalds 已提交
1527

1528
static struct static_key netstamp_needed __read_mostly;
1529
#ifdef HAVE_JUMP_LABEL
1530
/* We are not allowed to call static_key_slow_dec() from irq context
1531
 * If net_disable_timestamp() is called from irq context, defer the
1532
 * static_key_slow_dec() calls.
1533 1534 1535
 */
static atomic_t netstamp_needed_deferred;
#endif
L
Linus Torvalds 已提交
1536 1537 1538

void net_enable_timestamp(void)
{
1539 1540 1541 1542 1543
#ifdef HAVE_JUMP_LABEL
	int deferred = atomic_xchg(&netstamp_needed_deferred, 0);

	if (deferred) {
		while (--deferred)
1544
			static_key_slow_dec(&netstamp_needed);
1545 1546 1547 1548
		return;
	}
#endif
	WARN_ON(in_interrupt());
1549
	static_key_slow_inc(&netstamp_needed);
L
Linus Torvalds 已提交
1550
}
E
Eric Dumazet 已提交
1551
EXPORT_SYMBOL(net_enable_timestamp);
L
Linus Torvalds 已提交
1552 1553 1554

void net_disable_timestamp(void)
{
1555 1556 1557 1558 1559 1560
#ifdef HAVE_JUMP_LABEL
	if (in_interrupt()) {
		atomic_inc(&netstamp_needed_deferred);
		return;
	}
#endif
1561
	static_key_slow_dec(&netstamp_needed);
L
Linus Torvalds 已提交
1562
}
E
Eric Dumazet 已提交
1563
EXPORT_SYMBOL(net_disable_timestamp);
L
Linus Torvalds 已提交
1564

E
Eric Dumazet 已提交
1565
static inline void net_timestamp_set(struct sk_buff *skb)
L
Linus Torvalds 已提交
1566
{
1567
	skb->tstamp.tv64 = 0;
1568
	if (static_key_false(&netstamp_needed))
1569
		__net_timestamp(skb);
L
Linus Torvalds 已提交
1570 1571
}

1572
#define net_timestamp_check(COND, SKB)			\
1573
	if (static_key_false(&netstamp_needed)) {		\
1574 1575 1576
		if ((COND) && !(SKB)->tstamp.tv64)	\
			__net_timestamp(SKB);		\
	}						\
E
Eric Dumazet 已提交
1577

1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598
static inline bool is_skb_forwardable(struct net_device *dev,
				      struct sk_buff *skb)
{
	unsigned int len;

	if (!(dev->flags & IFF_UP))
		return false;

	len = dev->mtu + dev->hard_header_len + VLAN_HLEN;
	if (skb->len <= len)
		return true;

	/* if TSO is enabled, we don't care about the length as the packet
	 * could be forwarded without being segmented before
	 */
	if (skb_is_gso(skb))
		return true;

	return false;
}

1599 1600 1601 1602 1603 1604 1605 1606
/**
 * dev_forward_skb - loopback an skb to another netif
 *
 * @dev: destination network device
 * @skb: buffer to forward
 *
 * return values:
 *	NET_RX_SUCCESS	(no congestion)
1607
 *	NET_RX_DROP     (packet was dropped, but freed)
1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618
 *
 * dev_forward_skb can be used for injecting an skb from the
 * start_xmit function of one device into the receive queue
 * of another device.
 *
 * The receiving device may be in another namespace, so
 * we have to clear all information in the skb that could
 * impact namespace isolation.
 */
int dev_forward_skb(struct net_device *dev, struct sk_buff *skb)
{
1619 1620 1621 1622 1623 1624 1625 1626
	if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY) {
		if (skb_copy_ubufs(skb, GFP_ATOMIC)) {
			atomic_long_inc(&dev->rx_dropped);
			kfree_skb(skb);
			return NET_RX_DROP;
		}
	}

1627
	skb_orphan(skb);
1628
	nf_reset(skb);
1629

1630
	if (unlikely(!is_skb_forwardable(dev, skb))) {
1631
		atomic_long_inc(&dev->rx_dropped);
1632
		kfree_skb(skb);
1633
		return NET_RX_DROP;
1634
	}
1635
	skb->skb_iif = 0;
1636 1637
	skb->dev = dev;
	skb_dst_drop(skb);
1638 1639 1640
	skb->tstamp.tv64 = 0;
	skb->pkt_type = PACKET_HOST;
	skb->protocol = eth_type_trans(skb, dev);
1641 1642 1643
	skb->mark = 0;
	secpath_reset(skb);
	nf_reset(skb);
1644 1645 1646 1647
	return netif_rx(skb);
}
EXPORT_SYMBOL_GPL(dev_forward_skb);

1648 1649 1650 1651
static inline int deliver_skb(struct sk_buff *skb,
			      struct packet_type *pt_prev,
			      struct net_device *orig_dev)
{
1652 1653
	if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
		return -ENOMEM;
1654 1655 1656 1657
	atomic_inc(&skb->users);
	return pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
}

1658 1659
static inline bool skb_loop_sk(struct packet_type *ptype, struct sk_buff *skb)
{
1660
	if (!ptype->af_packet_priv || !skb->sk)
1661 1662 1663 1664 1665 1666 1667 1668 1669 1670
		return false;

	if (ptype->id_match)
		return ptype->id_match(ptype, skb->sk);
	else if ((struct sock *)ptype->af_packet_priv == skb->sk)
		return true;

	return false;
}

L
Linus Torvalds 已提交
1671 1672 1673 1674 1675
/*
 *	Support routine. Sends outgoing frames to any network
 *	taps currently in use.
 */

1676
static void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev)
L
Linus Torvalds 已提交
1677 1678
{
	struct packet_type *ptype;
1679 1680
	struct sk_buff *skb2 = NULL;
	struct packet_type *pt_prev = NULL;
1681

L
Linus Torvalds 已提交
1682 1683 1684 1685 1686 1687
	rcu_read_lock();
	list_for_each_entry_rcu(ptype, &ptype_all, list) {
		/* Never send packets back to the socket
		 * they originated from - MvS (miquels@drinkel.ow.org)
		 */
		if ((ptype->dev == dev || !ptype->dev) &&
1688
		    (!skb_loop_sk(ptype, skb))) {
1689 1690 1691 1692 1693 1694 1695
			if (pt_prev) {
				deliver_skb(skb2, pt_prev, skb->dev);
				pt_prev = ptype;
				continue;
			}

			skb2 = skb_clone(skb, GFP_ATOMIC);
L
Linus Torvalds 已提交
1696 1697 1698
			if (!skb2)
				break;

1699 1700
			net_timestamp_set(skb2);

L
Linus Torvalds 已提交
1701 1702 1703 1704
			/* skb->nh should be correctly
			   set by sender, so that the second statement is
			   just protection against buggy protocols.
			 */
1705
			skb_reset_mac_header(skb2);
L
Linus Torvalds 已提交
1706

1707
			if (skb_network_header(skb2) < skb2->data ||
1708
			    skb2->network_header > skb2->tail) {
1709 1710 1711
				net_crit_ratelimited("protocol %04x is buggy, dev %s\n",
						     ntohs(skb2->protocol),
						     dev->name);
1712
				skb_reset_network_header(skb2);
L
Linus Torvalds 已提交
1713 1714
			}

1715
			skb2->transport_header = skb2->network_header;
L
Linus Torvalds 已提交
1716
			skb2->pkt_type = PACKET_OUTGOING;
1717
			pt_prev = ptype;
L
Linus Torvalds 已提交
1718 1719
		}
	}
1720 1721
	if (pt_prev)
		pt_prev->func(skb2, skb->dev, pt_prev, skb->dev);
L
Linus Torvalds 已提交
1722 1723 1724
	rcu_read_unlock();
}

1725 1726
/**
 * netif_setup_tc - Handle tc mappings on real_num_tx_queues change
1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737
 * @dev: Network device
 * @txq: number of queues available
 *
 * If real_num_tx_queues is changed the tc mappings may no longer be
 * valid. To resolve this verify the tc mapping remains valid and if
 * not NULL the mapping. With no priorities mapping to this
 * offset/count pair it will no longer be used. In the worst case TC0
 * is invalid nothing can be done so disable priority mappings. If is
 * expected that drivers will fix this mapping if they can before
 * calling netif_set_real_num_tx_queues.
 */
E
Eric Dumazet 已提交
1738
static void netif_setup_tc(struct net_device *dev, unsigned int txq)
1739 1740 1741 1742 1743 1744
{
	int i;
	struct netdev_tc_txq *tc = &dev->tc_to_txq[0];

	/* If TC0 is invalidated disable TC mapping */
	if (tc->offset + tc->count > txq) {
1745
		pr_warn("Number of in use tx queues changed invalidating tc mappings. Priority traffic classification disabled!\n");
1746 1747 1748 1749 1750 1751 1752 1753 1754 1755
		dev->num_tc = 0;
		return;
	}

	/* Invalidated prio to tc mappings set to TC0 */
	for (i = 1; i < TC_BITMASK + 1; i++) {
		int q = netdev_get_prio_tc_map(dev, i);

		tc = &dev->tc_to_txq[q];
		if (tc->offset + tc->count > txq) {
1756 1757
			pr_warn("Number of in use tx queues changed. Priority %i to tc mapping %i is no longer valid. Setting map to 0\n",
				i, q);
1758 1759 1760 1761 1762
			netdev_set_prio_tc_map(dev, i, 0);
		}
	}
}

1763 1764 1765 1766 1767
#ifdef CONFIG_XPS
static DEFINE_MUTEX(xps_map_mutex);
#define xmap_dereference(P)		\
	rcu_dereference_protected((P), lockdep_is_held(&xps_map_mutex))

1768 1769
static struct xps_map *remove_xps_queue(struct xps_dev_maps *dev_maps,
					int cpu, u16 index)
1770
{
1771 1772
	struct xps_map *map = NULL;
	int pos;
1773

1774 1775
	if (dev_maps)
		map = xmap_dereference(dev_maps->cpu_map[cpu]);
1776

1777 1778
	for (pos = 0; map && pos < map->len; pos++) {
		if (map->queues[pos] == index) {
1779 1780 1781
			if (map->len > 1) {
				map->queues[pos] = map->queues[--map->len];
			} else {
1782
				RCU_INIT_POINTER(dev_maps->cpu_map[cpu], NULL);
1783 1784 1785
				kfree_rcu(map, rcu);
				map = NULL;
			}
1786
			break;
1787 1788 1789
		}
	}

1790 1791 1792
	return map;
}

1793
static void netif_reset_xps_queues_gt(struct net_device *dev, u16 index)
1794 1795
{
	struct xps_dev_maps *dev_maps;
1796
	int cpu, i;
1797 1798 1799 1800 1801 1802 1803 1804 1805
	bool active = false;

	mutex_lock(&xps_map_mutex);
	dev_maps = xmap_dereference(dev->xps_maps);

	if (!dev_maps)
		goto out_no_maps;

	for_each_possible_cpu(cpu) {
1806 1807 1808 1809 1810
		for (i = index; i < dev->num_tx_queues; i++) {
			if (!remove_xps_queue(dev_maps, cpu, i))
				break;
		}
		if (i == dev->num_tx_queues)
1811 1812 1813 1814
			active = true;
	}

	if (!active) {
1815 1816 1817 1818
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

1819 1820 1821 1822
	for (i = index; i < dev->num_tx_queues; i++)
		netdev_queue_numa_node_write(netdev_get_tx_queue(dev, i),
					     NUMA_NO_NODE);

1823 1824 1825 1826
out_no_maps:
	mutex_unlock(&xps_map_mutex);
}

1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861
static struct xps_map *expand_xps_map(struct xps_map *map,
				      int cpu, u16 index)
{
	struct xps_map *new_map;
	int alloc_len = XPS_MIN_MAP_ALLOC;
	int i, pos;

	for (pos = 0; map && pos < map->len; pos++) {
		if (map->queues[pos] != index)
			continue;
		return map;
	}

	/* Need to add queue to this CPU's existing map */
	if (map) {
		if (pos < map->alloc_len)
			return map;

		alloc_len = map->alloc_len * 2;
	}

	/* Need to allocate new map to store queue on this CPU's map */
	new_map = kzalloc_node(XPS_MAP_SIZE(alloc_len), GFP_KERNEL,
			       cpu_to_node(cpu));
	if (!new_map)
		return NULL;

	for (i = 0; i < pos; i++)
		new_map->queues[i] = map->queues[i];
	new_map->alloc_len = alloc_len;
	new_map->len = pos;

	return new_map;
}

1862 1863
int netif_set_xps_queue(struct net_device *dev, struct cpumask *mask, u16 index)
{
1864
	struct xps_dev_maps *dev_maps, *new_dev_maps = NULL;
1865 1866
	struct xps_map *map, *new_map;
	int maps_sz = max_t(unsigned int, XPS_DEV_MAPS_SIZE, L1_CACHE_BYTES);
1867 1868
	int cpu, numa_node_id = -2;
	bool active = false;
1869 1870 1871 1872 1873

	mutex_lock(&xps_map_mutex);

	dev_maps = xmap_dereference(dev->xps_maps);

1874 1875 1876 1877 1878 1879 1880
	/* allocate memory for queue storage */
	for_each_online_cpu(cpu) {
		if (!cpumask_test_cpu(cpu, mask))
			continue;

		if (!new_dev_maps)
			new_dev_maps = kzalloc(maps_sz, GFP_KERNEL);
1881 1882
		if (!new_dev_maps) {
			mutex_unlock(&xps_map_mutex);
1883
			return -ENOMEM;
1884
		}
1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898

		map = dev_maps ? xmap_dereference(dev_maps->cpu_map[cpu]) :
				 NULL;

		map = expand_xps_map(map, cpu, index);
		if (!map)
			goto error;

		RCU_INIT_POINTER(new_dev_maps->cpu_map[cpu], map);
	}

	if (!new_dev_maps)
		goto out_no_new_maps;

1899
	for_each_possible_cpu(cpu) {
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909
		if (cpumask_test_cpu(cpu, mask) && cpu_online(cpu)) {
			/* add queue to CPU maps */
			int pos = 0;

			map = xmap_dereference(new_dev_maps->cpu_map[cpu]);
			while ((pos < map->len) && (map->queues[pos] != index))
				pos++;

			if (pos == map->len)
				map->queues[map->len++] = index;
1910 1911 1912 1913 1914 1915
#ifdef CONFIG_NUMA
			if (numa_node_id == -2)
				numa_node_id = cpu_to_node(cpu);
			else if (numa_node_id != cpu_to_node(cpu))
				numa_node_id = -1;
#endif
1916 1917 1918 1919
		} else if (dev_maps) {
			/* fill in the new device map from the old device map */
			map = xmap_dereference(dev_maps->cpu_map[cpu]);
			RCU_INIT_POINTER(new_dev_maps->cpu_map[cpu], map);
1920
		}
1921

1922 1923
	}

1924 1925
	rcu_assign_pointer(dev->xps_maps, new_dev_maps);

1926
	/* Cleanup old maps */
1927 1928 1929 1930 1931 1932 1933
	if (dev_maps) {
		for_each_possible_cpu(cpu) {
			new_map = xmap_dereference(new_dev_maps->cpu_map[cpu]);
			map = xmap_dereference(dev_maps->cpu_map[cpu]);
			if (map && map != new_map)
				kfree_rcu(map, rcu);
		}
1934

1935
		kfree_rcu(dev_maps, rcu);
1936 1937
	}

1938 1939
	dev_maps = new_dev_maps;
	active = true;
1940

1941 1942
out_no_new_maps:
	/* update Tx queue numa node */
1943 1944 1945 1946
	netdev_queue_numa_node_write(netdev_get_tx_queue(dev, index),
				     (numa_node_id >= 0) ? numa_node_id :
				     NUMA_NO_NODE);

1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965
	if (!dev_maps)
		goto out_no_maps;

	/* removes queue from unused CPUs */
	for_each_possible_cpu(cpu) {
		if (cpumask_test_cpu(cpu, mask) && cpu_online(cpu))
			continue;

		if (remove_xps_queue(dev_maps, cpu, index))
			active = true;
	}

	/* free map if not active */
	if (!active) {
		RCU_INIT_POINTER(dev->xps_maps, NULL);
		kfree_rcu(dev_maps, rcu);
	}

out_no_maps:
1966 1967 1968 1969
	mutex_unlock(&xps_map_mutex);

	return 0;
error:
1970 1971 1972 1973 1974 1975 1976 1977 1978
	/* remove any maps that we added */
	for_each_possible_cpu(cpu) {
		new_map = xmap_dereference(new_dev_maps->cpu_map[cpu]);
		map = dev_maps ? xmap_dereference(dev_maps->cpu_map[cpu]) :
				 NULL;
		if (new_map && new_map != map)
			kfree(new_map);
	}

1979 1980 1981 1982 1983 1984 1985 1986
	mutex_unlock(&xps_map_mutex);

	kfree(new_dev_maps);
	return -ENOMEM;
}
EXPORT_SYMBOL(netif_set_xps_queue);

#endif
1987 1988 1989 1990
/*
 * Routine to help set real_num_tx_queues. To avoid skbs mapped to queues
 * greater then real_num_tx_queues stale skbs on the qdisc must be flushed.
 */
1991
int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq)
1992
{
T
Tom Herbert 已提交
1993 1994
	int rc;

1995 1996
	if (txq < 1 || txq > dev->num_tx_queues)
		return -EINVAL;
1997

1998 1999
	if (dev->reg_state == NETREG_REGISTERED ||
	    dev->reg_state == NETREG_UNREGISTERING) {
2000 2001
		ASSERT_RTNL();

T
Tom Herbert 已提交
2002 2003
		rc = netdev_queue_update_kobjects(dev, dev->real_num_tx_queues,
						  txq);
T
Tom Herbert 已提交
2004 2005 2006
		if (rc)
			return rc;

2007 2008 2009
		if (dev->num_tc)
			netif_setup_tc(dev, txq);

2010
		if (txq < dev->real_num_tx_queues) {
2011
			qdisc_reset_all_tx_gt(dev, txq);
2012 2013 2014 2015
#ifdef CONFIG_XPS
			netif_reset_xps_queues_gt(dev, txq);
#endif
		}
2016
	}
2017 2018 2019

	dev->real_num_tx_queues = txq;
	return 0;
2020 2021
}
EXPORT_SYMBOL(netif_set_real_num_tx_queues);
2022

2023 2024 2025 2026 2027 2028 2029 2030
#ifdef CONFIG_RPS
/**
 *	netif_set_real_num_rx_queues - set actual number of RX queues used
 *	@dev: Network device
 *	@rxq: Actual number of RX queues
 *
 *	This must be called either with the rtnl_lock held or before
 *	registration of the net device.  Returns 0 on success, or a
2031 2032
 *	negative error code.  If called before registration, it always
 *	succeeds.
2033 2034 2035 2036 2037
 */
int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
{
	int rc;

T
Tom Herbert 已提交
2038 2039 2040
	if (rxq < 1 || rxq > dev->num_rx_queues)
		return -EINVAL;

2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
	if (dev->reg_state == NETREG_REGISTERED) {
		ASSERT_RTNL();

		rc = net_rx_queue_update_kobjects(dev, dev->real_num_rx_queues,
						  rxq);
		if (rc)
			return rc;
	}

	dev->real_num_rx_queues = rxq;
	return 0;
}
EXPORT_SYMBOL(netif_set_real_num_rx_queues);
#endif

2056 2057
/**
 * netif_get_num_default_rss_queues - default number of RSS queues
2058 2059 2060 2061
 *
 * This routine should set an upper limit on the number of RSS queues
 * used by default by multiqueue devices.
 */
2062
int netif_get_num_default_rss_queues(void)
2063 2064 2065 2066 2067
{
	return min_t(int, DEFAULT_MAX_NUM_RSS_QUEUES, num_online_cpus());
}
EXPORT_SYMBOL(netif_get_num_default_rss_queues);

2068
static inline void __netif_reschedule(struct Qdisc *q)
2069
{
2070 2071
	struct softnet_data *sd;
	unsigned long flags;
2072

2073 2074
	local_irq_save(flags);
	sd = &__get_cpu_var(softnet_data);
2075 2076 2077
	q->next_sched = NULL;
	*sd->output_queue_tailp = q;
	sd->output_queue_tailp = &q->next_sched;
2078 2079 2080 2081 2082 2083 2084 2085
	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_restore(flags);
}

void __netif_schedule(struct Qdisc *q)
{
	if (!test_and_set_bit(__QDISC_STATE_SCHED, &q->state))
		__netif_reschedule(q);
2086 2087 2088
}
EXPORT_SYMBOL(__netif_schedule);

2089
void dev_kfree_skb_irq(struct sk_buff *skb)
2090
{
2091
	if (atomic_dec_and_test(&skb->users)) {
2092 2093
		struct softnet_data *sd;
		unsigned long flags;
2094

2095 2096 2097 2098 2099 2100 2101
		local_irq_save(flags);
		sd = &__get_cpu_var(softnet_data);
		skb->next = sd->completion_queue;
		sd->completion_queue = skb;
		raise_softirq_irqoff(NET_TX_SOFTIRQ);
		local_irq_restore(flags);
	}
2102
}
2103
EXPORT_SYMBOL(dev_kfree_skb_irq);
2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114

void dev_kfree_skb_any(struct sk_buff *skb)
{
	if (in_irq() || irqs_disabled())
		dev_kfree_skb_irq(skb);
	else
		dev_kfree_skb(skb);
}
EXPORT_SYMBOL(dev_kfree_skb_any);


2115 2116 2117 2118 2119 2120
/**
 * netif_device_detach - mark device as removed
 * @dev: network device
 *
 * Mark device as removed from system and therefore no longer available.
 */
2121 2122 2123 2124
void netif_device_detach(struct net_device *dev)
{
	if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2125
		netif_tx_stop_all_queues(dev);
2126 2127 2128 2129
	}
}
EXPORT_SYMBOL(netif_device_detach);

2130 2131 2132 2133 2134 2135
/**
 * netif_device_attach - mark device as attached
 * @dev: network device
 *
 * Mark device as attached from system and restart if needed.
 */
2136 2137 2138 2139
void netif_device_attach(struct net_device *dev)
{
	if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
	    netif_running(dev)) {
2140
		netif_tx_wake_all_queues(dev);
2141
		__netdev_watchdog_up(dev);
2142 2143 2144 2145
	}
}
EXPORT_SYMBOL(netif_device_attach);

2146 2147
static void skb_warn_bad_offload(const struct sk_buff *skb)
{
2148
	static const netdev_features_t null_features = 0;
2149 2150 2151 2152 2153 2154 2155 2156
	struct net_device *dev = skb->dev;
	const char *driver = "";

	if (dev && dev->dev.parent)
		driver = dev_driver_string(dev->dev.parent);

	WARN(1, "%s: caps=(%pNF, %pNF) len=%d data_len=%d gso_size=%d "
	     "gso_type=%d ip_summed=%d\n",
2157 2158
	     driver, dev ? &dev->features : &null_features,
	     skb->sk ? &skb->sk->sk_route_caps : &null_features,
2159 2160 2161 2162
	     skb->len, skb->data_len, skb_shinfo(skb)->gso_size,
	     skb_shinfo(skb)->gso_type, skb->ip_summed);
}

L
Linus Torvalds 已提交
2163 2164 2165 2166
/*
 * Invalidate hardware checksum when packet is to be mangled, and
 * complete checksum manually on outgoing path.
 */
2167
int skb_checksum_help(struct sk_buff *skb)
L
Linus Torvalds 已提交
2168
{
2169
	__wsum csum;
2170
	int ret = 0, offset;
L
Linus Torvalds 已提交
2171

2172
	if (skb->ip_summed == CHECKSUM_COMPLETE)
2173 2174 2175
		goto out_set_summed;

	if (unlikely(skb_shinfo(skb)->gso_size)) {
2176 2177
		skb_warn_bad_offload(skb);
		return -EINVAL;
L
Linus Torvalds 已提交
2178 2179
	}

2180 2181 2182 2183 2184 2185 2186 2187 2188
	/* Before computing a checksum, we should make sure no frag could
	 * be modified by an external entity : checksum could be wrong.
	 */
	if (skb_has_shared_frag(skb)) {
		ret = __skb_linearize(skb);
		if (ret)
			goto out;
	}

2189
	offset = skb_checksum_start_offset(skb);
2190 2191 2192 2193 2194 2195 2196 2197
	BUG_ON(offset >= skb_headlen(skb));
	csum = skb_checksum(skb, offset, skb->len - offset, 0);

	offset += skb->csum_offset;
	BUG_ON(offset + sizeof(__sum16) > skb_headlen(skb));

	if (skb_cloned(skb) &&
	    !skb_clone_writable(skb, offset + sizeof(__sum16))) {
L
Linus Torvalds 已提交
2198 2199 2200 2201 2202
		ret = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
		if (ret)
			goto out;
	}

2203
	*(__sum16 *)(skb->data + offset) = csum_fold(csum);
2204
out_set_summed:
L
Linus Torvalds 已提交
2205
	skb->ip_summed = CHECKSUM_NONE;
2206
out:
L
Linus Torvalds 已提交
2207 2208
	return ret;
}
E
Eric Dumazet 已提交
2209
EXPORT_SYMBOL(skb_checksum_help);
L
Linus Torvalds 已提交
2210

2211
/**
2212
 *	skb_mac_gso_segment - mac layer segmentation handler.
2213
 *	@skb: buffer to segment
2214
 *	@features: features for the output path (see dev->features)
2215
 */
2216 2217
struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
				    netdev_features_t features)
2218 2219
{
	struct sk_buff *segs = ERR_PTR(-EPROTONOSUPPORT);
2220
	struct packet_offload *ptype;
A
Al Viro 已提交
2221
	__be16 type = skb->protocol;
2222

2223
	while (type == htons(ETH_P_8021Q)) {
2224
		int vlan_depth = ETH_HLEN;
2225
		struct vlan_hdr *vh;
2226

2227
		if (unlikely(!pskb_may_pull(skb, vlan_depth + VLAN_HLEN)))
2228 2229
			return ERR_PTR(-EINVAL);

2230 2231 2232
		vh = (struct vlan_hdr *)(skb->data + vlan_depth);
		type = vh->h_vlan_encapsulated_proto;
		vlan_depth += VLAN_HLEN;
2233 2234
	}

2235 2236 2237
	__skb_pull(skb, skb->mac_len);

	rcu_read_lock();
2238
	list_for_each_entry_rcu(ptype, &offload_base, list) {
2239
		if (ptype->type == type && ptype->callbacks.gso_segment) {
2240
			if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL)) {
2241 2242
				int err;

2243
				err = ptype->callbacks.gso_send_check(skb);
2244 2245 2246
				segs = ERR_PTR(err);
				if (err || skb_gso_ok(skb, features))
					break;
2247 2248
				__skb_push(skb, (skb->data -
						 skb_network_header(skb)));
2249
			}
2250
			segs = ptype->callbacks.gso_segment(skb, features);
2251 2252 2253 2254 2255
			break;
		}
	}
	rcu_read_unlock();

2256
	__skb_push(skb, skb->data - skb_mac_header(skb));
2257

2258 2259
	return segs;
}
2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296
EXPORT_SYMBOL(skb_mac_gso_segment);


/* openvswitch calls this on rx path, so we need a different check.
 */
static inline bool skb_needs_check(struct sk_buff *skb, bool tx_path)
{
	if (tx_path)
		return skb->ip_summed != CHECKSUM_PARTIAL;
	else
		return skb->ip_summed == CHECKSUM_NONE;
}

/**
 *	__skb_gso_segment - Perform segmentation on skb.
 *	@skb: buffer to segment
 *	@features: features for the output path (see dev->features)
 *	@tx_path: whether it is called in TX path
 *
 *	This function segments the given skb and returns a list of segments.
 *
 *	It may return NULL if the skb requires no segmentation.  This is
 *	only possible when GSO is used for verifying header integrity.
 */
struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
				  netdev_features_t features, bool tx_path)
{
	if (unlikely(skb_needs_check(skb, tx_path))) {
		int err;

		skb_warn_bad_offload(skb);

		if (skb_header_cloned(skb) &&
		    (err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC)))
			return ERR_PTR(err);
	}

2297
	SKB_GSO_CB(skb)->mac_offset = skb_headroom(skb);
2298 2299 2300 2301 2302
	skb_reset_mac_header(skb);
	skb_reset_mac_len(skb);

	return skb_mac_gso_segment(skb, features);
}
2303
EXPORT_SYMBOL(__skb_gso_segment);
2304

2305 2306 2307 2308 2309
/* Take action when hardware reception checksum errors are detected. */
#ifdef CONFIG_BUG
void netdev_rx_csum_fault(struct net_device *dev)
{
	if (net_ratelimit()) {
2310
		pr_err("%s: hw csum failure\n", dev ? dev->name : "<unknown>");
2311 2312 2313 2314 2315 2316
		dump_stack();
	}
}
EXPORT_SYMBOL(netdev_rx_csum_fault);
#endif

L
Linus Torvalds 已提交
2317 2318 2319 2320 2321
/* Actually, we should eliminate this check as soon as we know, that:
 * 1. IOMMU is present and allows to map all the memory.
 * 2. No high memory really exists on this machine.
 */

E
Eric Dumazet 已提交
2322
static int illegal_highdma(struct net_device *dev, struct sk_buff *skb)
L
Linus Torvalds 已提交
2323
{
2324
#ifdef CONFIG_HIGHMEM
L
Linus Torvalds 已提交
2325
	int i;
2326
	if (!(dev->features & NETIF_F_HIGHDMA)) {
2327 2328 2329
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			if (PageHighMem(skb_frag_page(frag)))
2330
				return 1;
2331
		}
2332
	}
L
Linus Torvalds 已提交
2333

2334 2335
	if (PCI_DMA_BUS_IS_PHYS) {
		struct device *pdev = dev->dev.parent;
L
Linus Torvalds 已提交
2336

E
Eric Dumazet 已提交
2337 2338
		if (!pdev)
			return 0;
2339
		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
2340 2341
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			dma_addr_t addr = page_to_phys(skb_frag_page(frag));
2342 2343 2344 2345
			if (!pdev->dma_mask || addr + PAGE_SIZE - 1 > *pdev->dma_mask)
				return 1;
		}
	}
2346
#endif
L
Linus Torvalds 已提交
2347 2348 2349
	return 0;
}

2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375
struct dev_gso_cb {
	void (*destructor)(struct sk_buff *skb);
};

#define DEV_GSO_CB(skb) ((struct dev_gso_cb *)(skb)->cb)

static void dev_gso_skb_destructor(struct sk_buff *skb)
{
	struct dev_gso_cb *cb;

	do {
		struct sk_buff *nskb = skb->next;

		skb->next = nskb->next;
		nskb->next = NULL;
		kfree_skb(nskb);
	} while (skb->next);

	cb = DEV_GSO_CB(skb);
	if (cb->destructor)
		cb->destructor(skb);
}

/**
 *	dev_gso_segment - Perform emulated hardware segmentation on skb.
 *	@skb: buffer to segment
2376
 *	@features: device features as applicable to this skb
2377 2378 2379 2380
 *
 *	This function segments the given skb and stores the list of segments
 *	in skb->next.
 */
2381
static int dev_gso_segment(struct sk_buff *skb, netdev_features_t features)
2382 2383
{
	struct sk_buff *segs;
2384 2385 2386 2387 2388 2389

	segs = skb_gso_segment(skb, features);

	/* Verifying header integrity only. */
	if (!segs)
		return 0;
2390

2391
	if (IS_ERR(segs))
2392 2393 2394 2395 2396 2397 2398 2399 2400
		return PTR_ERR(segs);

	skb->next = segs;
	DEV_GSO_CB(skb)->destructor = skb->destructor;
	skb->destructor = dev_gso_skb_destructor;

	return 0;
}

2401
static bool can_checksum_protocol(netdev_features_t features, __be16 protocol)
2402 2403 2404 2405 2406 2407 2408 2409 2410 2411
{
	return ((features & NETIF_F_GEN_CSUM) ||
		((features & NETIF_F_V4_CSUM) &&
		 protocol == htons(ETH_P_IP)) ||
		((features & NETIF_F_V6_CSUM) &&
		 protocol == htons(ETH_P_IPV6)) ||
		((features & NETIF_F_FCOE_CRC) &&
		 protocol == htons(ETH_P_FCOE)));
}

2412 2413
static netdev_features_t harmonize_features(struct sk_buff *skb,
	__be16 protocol, netdev_features_t features)
2414
{
2415 2416
	if (skb->ip_summed != CHECKSUM_NONE &&
	    !can_checksum_protocol(features, protocol)) {
2417 2418 2419 2420 2421 2422 2423 2424 2425
		features &= ~NETIF_F_ALL_CSUM;
		features &= ~NETIF_F_SG;
	} else if (illegal_highdma(skb->dev, skb)) {
		features &= ~NETIF_F_SG;
	}

	return features;
}

2426
netdev_features_t netif_skb_features(struct sk_buff *skb)
2427 2428
{
	__be16 protocol = skb->protocol;
2429
	netdev_features_t features = skb->dev->features;
2430

2431 2432 2433
	if (skb_shinfo(skb)->gso_segs > skb->dev->gso_max_segs)
		features &= ~NETIF_F_GSO_MASK;

2434 2435 2436
	if (protocol == htons(ETH_P_8021Q)) {
		struct vlan_ethhdr *veh = (struct vlan_ethhdr *)skb->data;
		protocol = veh->h_vlan_encapsulated_proto;
2437 2438 2439
	} else if (!vlan_tx_tag_present(skb)) {
		return harmonize_features(skb, protocol, features);
	}
2440

2441
	features &= (skb->dev->vlan_features | NETIF_F_HW_VLAN_TX);
2442 2443 2444 2445 2446

	if (protocol != htons(ETH_P_8021Q)) {
		return harmonize_features(skb, protocol, features);
	} else {
		features &= NETIF_F_SG | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST |
2447
				NETIF_F_GEN_CSUM | NETIF_F_HW_VLAN_TX;
2448 2449
		return harmonize_features(skb, protocol, features);
	}
2450
}
2451
EXPORT_SYMBOL(netif_skb_features);
2452

2453 2454 2455
/*
 * Returns true if either:
 *	1. skb has frag_list and the device doesn't support FRAGLIST, or
2456
 *	2. skb is fragmented and the device does not support SG.
2457 2458
 */
static inline int skb_needs_linearize(struct sk_buff *skb,
2459
				      int features)
2460
{
2461 2462 2463
	return skb_is_nonlinear(skb) &&
			((skb_has_frag_list(skb) &&
				!(features & NETIF_F_FRAGLIST)) ||
2464
			(skb_shinfo(skb)->nr_frags &&
2465
				!(features & NETIF_F_SG)));
2466 2467
}

2468 2469
int dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
			struct netdev_queue *txq)
2470
{
2471
	const struct net_device_ops *ops = dev->netdev_ops;
2472
	int rc = NETDEV_TX_OK;
2473
	unsigned int skb_len;
2474

2475
	if (likely(!skb->next)) {
2476
		netdev_features_t features;
2477

2478
		/*
L
Lucas De Marchi 已提交
2479
		 * If device doesn't need skb->dst, release it right now while
2480 2481
		 * its hot in this cpu cache
		 */
E
Eric Dumazet 已提交
2482 2483 2484
		if (dev->priv_flags & IFF_XMIT_DST_RELEASE)
			skb_dst_drop(skb);

2485 2486
		features = netif_skb_features(skb);

2487
		if (vlan_tx_tag_present(skb) &&
2488
		    !(features & NETIF_F_HW_VLAN_TX)) {
2489 2490 2491 2492 2493 2494 2495
			skb = __vlan_put_tag(skb, vlan_tx_tag_get(skb));
			if (unlikely(!skb))
				goto out;

			skb->vlan_tci = 0;
		}

2496 2497 2498 2499 2500 2501 2502
		/* If encapsulation offload request, verify we are testing
		 * hardware encapsulation features instead of standard
		 * features for the netdev
		 */
		if (skb->encapsulation)
			features &= dev->hw_enc_features;

2503
		if (netif_needs_gso(skb, features)) {
2504
			if (unlikely(dev_gso_segment(skb, features)))
2505 2506 2507
				goto out_kfree_skb;
			if (skb->next)
				goto gso;
2508
		} else {
2509
			if (skb_needs_linearize(skb, features) &&
2510 2511 2512 2513 2514 2515 2516 2517
			    __skb_linearize(skb))
				goto out_kfree_skb;

			/* If packet is not checksummed and device does not
			 * support checksumming for this protocol, complete
			 * checksumming here.
			 */
			if (skb->ip_summed == CHECKSUM_PARTIAL) {
2518 2519 2520 2521 2522 2523
				if (skb->encapsulation)
					skb_set_inner_transport_header(skb,
						skb_checksum_start_offset(skb));
				else
					skb_set_transport_header(skb,
						skb_checksum_start_offset(skb));
2524
				if (!(features & NETIF_F_ALL_CSUM) &&
2525 2526 2527
				     skb_checksum_help(skb))
					goto out_kfree_skb;
			}
2528 2529
		}

2530 2531 2532
		if (!list_empty(&ptype_all))
			dev_queue_xmit_nit(skb, dev);

2533
		skb_len = skb->len;
2534
		rc = ops->ndo_start_xmit(skb, dev);
2535
		trace_net_dev_xmit(skb, rc, dev, skb_len);
2536
		if (rc == NETDEV_TX_OK)
E
Eric Dumazet 已提交
2537
			txq_trans_update(txq);
2538
		return rc;
2539 2540
	}

2541
gso:
2542 2543 2544 2545 2546
	do {
		struct sk_buff *nskb = skb->next;

		skb->next = nskb->next;
		nskb->next = NULL;
2547 2548

		/*
L
Lucas De Marchi 已提交
2549
		 * If device doesn't need nskb->dst, release it right now while
2550 2551 2552 2553 2554
		 * its hot in this cpu cache
		 */
		if (dev->priv_flags & IFF_XMIT_DST_RELEASE)
			skb_dst_drop(nskb);

2555 2556 2557
		if (!list_empty(&ptype_all))
			dev_queue_xmit_nit(nskb, dev);

2558
		skb_len = nskb->len;
2559
		rc = ops->ndo_start_xmit(nskb, dev);
2560
		trace_net_dev_xmit(nskb, rc, dev, skb_len);
2561
		if (unlikely(rc != NETDEV_TX_OK)) {
2562 2563
			if (rc & ~NETDEV_TX_MASK)
				goto out_kfree_gso_skb;
2564
			nskb->next = skb->next;
2565 2566 2567
			skb->next = nskb;
			return rc;
		}
E
Eric Dumazet 已提交
2568
		txq_trans_update(txq);
2569
		if (unlikely(netif_xmit_stopped(txq) && skb->next))
2570
			return NETDEV_TX_BUSY;
2571
	} while (skb->next);
2572

2573 2574 2575
out_kfree_gso_skb:
	if (likely(skb->next == NULL))
		skb->destructor = DEV_GSO_CB(skb)->destructor;
2576 2577
out_kfree_skb:
	kfree_skb(skb);
2578
out:
2579
	return rc;
2580 2581
}

2582 2583 2584 2585 2586 2587 2588 2589 2590 2591
static void qdisc_pkt_len_init(struct sk_buff *skb)
{
	const struct skb_shared_info *shinfo = skb_shinfo(skb);

	qdisc_skb_cb(skb)->pkt_len = skb->len;

	/* To get more precise estimation of bytes sent on wire,
	 * we add to pkt_len the headers size of all segments
	 */
	if (shinfo->gso_size)  {
2592
		unsigned int hdr_len;
2593

2594 2595 2596 2597
		/* mac layer + network layer */
		hdr_len = skb_transport_header(skb) - skb_mac_header(skb);

		/* + transport layer */
2598 2599 2600 2601 2602 2603 2604 2605
		if (likely(shinfo->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6)))
			hdr_len += tcp_hdrlen(skb);
		else
			hdr_len += sizeof(struct udphdr);
		qdisc_skb_cb(skb)->pkt_len += (shinfo->gso_segs - 1) * hdr_len;
	}
}

2606 2607 2608 2609 2610
static inline int __dev_xmit_skb(struct sk_buff *skb, struct Qdisc *q,
				 struct net_device *dev,
				 struct netdev_queue *txq)
{
	spinlock_t *root_lock = qdisc_lock(q);
E
Eric Dumazet 已提交
2611
	bool contended;
2612 2613
	int rc;

2614
	qdisc_pkt_len_init(skb);
E
Eric Dumazet 已提交
2615
	qdisc_calculate_pkt_len(skb, q);
2616 2617 2618 2619 2620 2621
	/*
	 * Heuristic to force contended enqueues to serialize on a
	 * separate lock before trying to get qdisc main lock.
	 * This permits __QDISC_STATE_RUNNING owner to get the lock more often
	 * and dequeue packets faster.
	 */
E
Eric Dumazet 已提交
2622
	contended = qdisc_is_running(q);
2623 2624 2625
	if (unlikely(contended))
		spin_lock(&q->busylock);

2626 2627 2628 2629 2630
	spin_lock(root_lock);
	if (unlikely(test_bit(__QDISC_STATE_DEACTIVATED, &q->state))) {
		kfree_skb(skb);
		rc = NET_XMIT_DROP;
	} else if ((q->flags & TCQ_F_CAN_BYPASS) && !qdisc_qlen(q) &&
2631
		   qdisc_run_begin(q)) {
2632 2633 2634 2635 2636
		/*
		 * This is a work-conserving queue; there are no old skbs
		 * waiting to be sent out; and the qdisc is not running -
		 * xmit the skb directly.
		 */
E
Eric Dumazet 已提交
2637 2638
		if (!(dev->priv_flags & IFF_XMIT_DST_RELEASE))
			skb_dst_force(skb);
2639 2640 2641

		qdisc_bstats_update(q, skb);

2642 2643 2644 2645 2646
		if (sch_direct_xmit(skb, q, dev, txq, root_lock)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
2647
			__qdisc_run(q);
2648
		} else
2649
			qdisc_run_end(q);
2650 2651 2652

		rc = NET_XMIT_SUCCESS;
	} else {
E
Eric Dumazet 已提交
2653
		skb_dst_force(skb);
E
Eric Dumazet 已提交
2654
		rc = q->enqueue(skb, q) & NET_XMIT_MASK;
2655 2656 2657 2658 2659 2660 2661
		if (qdisc_run_begin(q)) {
			if (unlikely(contended)) {
				spin_unlock(&q->busylock);
				contended = false;
			}
			__qdisc_run(q);
		}
2662 2663
	}
	spin_unlock(root_lock);
2664 2665
	if (unlikely(contended))
		spin_unlock(&q->busylock);
2666 2667 2668
	return rc;
}

2669 2670 2671
#if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
static void skb_update_prio(struct sk_buff *skb)
{
I
Igor Maravic 已提交
2672
	struct netprio_map *map = rcu_dereference_bh(skb->dev->priomap);
2673

2674 2675 2676 2677 2678 2679
	if (!skb->priority && skb->sk && map) {
		unsigned int prioidx = skb->sk->sk_cgrp_prioidx;

		if (prioidx < map->priomap_len)
			skb->priority = map->priomap[prioidx];
	}
2680 2681 2682 2683 2684
}
#else
#define skb_update_prio(skb)
#endif

2685
static DEFINE_PER_CPU(int, xmit_recursion);
2686
#define RECURSION_LIMIT 10
2687

2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704
/**
 *	dev_loopback_xmit - loop back @skb
 *	@skb: buffer to transmit
 */
int dev_loopback_xmit(struct sk_buff *skb)
{
	skb_reset_mac_header(skb);
	__skb_pull(skb, skb_network_offset(skb));
	skb->pkt_type = PACKET_LOOPBACK;
	skb->ip_summed = CHECKSUM_UNNECESSARY;
	WARN_ON(!skb_dst(skb));
	skb_dst_force(skb);
	netif_rx_ni(skb);
	return 0;
}
EXPORT_SYMBOL(dev_loopback_xmit);

2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729
/**
 *	dev_queue_xmit - transmit a buffer
 *	@skb: buffer to transmit
 *
 *	Queue a buffer for transmission to a network device. The caller must
 *	have set the device and priority and built the buffer before calling
 *	this function. The function can be called from an interrupt.
 *
 *	A negative errno code is returned on a failure. A success does not
 *	guarantee the frame will be transmitted as it may be dropped due
 *	to congestion or traffic shaping.
 *
 * -----------------------------------------------------------------------------------
 *      I notice this method can also return errors from the queue disciplines,
 *      including NET_XMIT_DROP, which is a positive value.  So, errors can also
 *      be positive.
 *
 *      Regardless of the return value, the skb is consumed, so it is currently
 *      difficult to retry a send to this method.  (You can bump the ref count
 *      before sending to hold a reference for retry if you are careful.)
 *
 *      When calling this method, interrupts MUST be enabled.  This is because
 *      the BH enable code must have IRQs enabled so that it will not deadlock.
 *          --BLG
 */
L
Linus Torvalds 已提交
2730 2731 2732
int dev_queue_xmit(struct sk_buff *skb)
{
	struct net_device *dev = skb->dev;
2733
	struct netdev_queue *txq;
L
Linus Torvalds 已提交
2734 2735 2736
	struct Qdisc *q;
	int rc = -ENOMEM;

2737 2738
	skb_reset_mac_header(skb);

2739 2740
	/* Disable soft irqs for various locks below. Also
	 * stops preemption for RCU.
L
Linus Torvalds 已提交
2741
	 */
2742
	rcu_read_lock_bh();
L
Linus Torvalds 已提交
2743

2744 2745
	skb_update_prio(skb);

2746
	txq = netdev_pick_tx(dev, skb);
2747
	q = rcu_dereference_bh(txq->qdisc);
2748

L
Linus Torvalds 已提交
2749
#ifdef CONFIG_NET_CLS_ACT
E
Eric Dumazet 已提交
2750
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_EGRESS);
L
Linus Torvalds 已提交
2751
#endif
2752
	trace_net_dev_queue(skb);
L
Linus Torvalds 已提交
2753
	if (q->enqueue) {
2754
		rc = __dev_xmit_skb(skb, q, dev, txq);
2755
		goto out;
L
Linus Torvalds 已提交
2756 2757 2758 2759 2760
	}

	/* The device has no queue. Common case for software devices:
	   loopback, all the sorts of tunnels...

H
Herbert Xu 已提交
2761 2762
	   Really, it is unlikely that netif_tx_lock protection is necessary
	   here.  (f.e. loopback and IP tunnels are clean ignoring statistics
L
Linus Torvalds 已提交
2763 2764 2765 2766 2767 2768 2769 2770 2771 2772
	   counters.)
	   However, it is possible, that they rely on protection
	   made by us here.

	   Check this and shot the lock. It is not prone from deadlocks.
	   Either shot noqueue qdisc, it is even simpler 8)
	 */
	if (dev->flags & IFF_UP) {
		int cpu = smp_processor_id(); /* ok because BHs are off */

2773
		if (txq->xmit_lock_owner != cpu) {
L
Linus Torvalds 已提交
2774

2775 2776 2777
			if (__this_cpu_read(xmit_recursion) > RECURSION_LIMIT)
				goto recursion_alert;

2778
			HARD_TX_LOCK(dev, txq, cpu);
L
Linus Torvalds 已提交
2779

2780
			if (!netif_xmit_stopped(txq)) {
2781
				__this_cpu_inc(xmit_recursion);
2782
				rc = dev_hard_start_xmit(skb, dev, txq);
2783
				__this_cpu_dec(xmit_recursion);
2784
				if (dev_xmit_complete(rc)) {
2785
					HARD_TX_UNLOCK(dev, txq);
L
Linus Torvalds 已提交
2786 2787 2788
					goto out;
				}
			}
2789
			HARD_TX_UNLOCK(dev, txq);
2790 2791
			net_crit_ratelimited("Virtual device %s asks to queue packet!\n",
					     dev->name);
L
Linus Torvalds 已提交
2792 2793
		} else {
			/* Recursion is detected! It is possible,
2794 2795 2796
			 * unfortunately
			 */
recursion_alert:
2797 2798
			net_crit_ratelimited("Dead loop on virtual device %s, fix it urgently!\n",
					     dev->name);
L
Linus Torvalds 已提交
2799 2800 2801 2802
		}
	}

	rc = -ENETDOWN;
2803
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
2804 2805 2806 2807

	kfree_skb(skb);
	return rc;
out:
2808
	rcu_read_unlock_bh();
L
Linus Torvalds 已提交
2809 2810
	return rc;
}
E
Eric Dumazet 已提交
2811
EXPORT_SYMBOL(dev_queue_xmit);
L
Linus Torvalds 已提交
2812 2813 2814 2815 2816 2817


/*=======================================================================
			Receiver routines
  =======================================================================*/

2818
int netdev_max_backlog __read_mostly = 1000;
E
Eric Dumazet 已提交
2819 2820
EXPORT_SYMBOL(netdev_max_backlog);

E
Eric Dumazet 已提交
2821
int netdev_tstamp_prequeue __read_mostly = 1;
2822 2823
int netdev_budget __read_mostly = 300;
int weight_p __read_mostly = 64;            /* old backlog weight */
L
Linus Torvalds 已提交
2824

E
Eric Dumazet 已提交
2825 2826 2827 2828 2829 2830 2831 2832
/* Called with irq disabled */
static inline void ____napi_schedule(struct softnet_data *sd,
				     struct napi_struct *napi)
{
	list_add_tail(&napi->poll_list, &sd->poll_list);
	__raise_softirq_irqoff(NET_RX_SOFTIRQ);
}

2833 2834 2835
#ifdef CONFIG_RPS

/* One global table that all flow-based protocols share. */
E
Eric Dumazet 已提交
2836
struct rps_sock_flow_table __rcu *rps_sock_flow_table __read_mostly;
2837 2838
EXPORT_SYMBOL(rps_sock_flow_table);

2839
struct static_key rps_needed __read_mostly;
2840

2841 2842 2843 2844
static struct rps_dev_flow *
set_rps_cpu(struct net_device *dev, struct sk_buff *skb,
	    struct rps_dev_flow *rflow, u16 next_cpu)
{
2845
	if (next_cpu != RPS_NO_CPU) {
2846 2847 2848 2849 2850 2851 2852 2853 2854
#ifdef CONFIG_RFS_ACCEL
		struct netdev_rx_queue *rxqueue;
		struct rps_dev_flow_table *flow_table;
		struct rps_dev_flow *old_rflow;
		u32 flow_id;
		u16 rxq_index;
		int rc;

		/* Should we steer this flow to a different hardware queue? */
2855 2856
		if (!skb_rx_queue_recorded(skb) || !dev->rx_cpu_rmap ||
		    !(dev->features & NETIF_F_NTUPLE))
2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878
			goto out;
		rxq_index = cpu_rmap_lookup_index(dev->rx_cpu_rmap, next_cpu);
		if (rxq_index == skb_get_rx_queue(skb))
			goto out;

		rxqueue = dev->_rx + rxq_index;
		flow_table = rcu_dereference(rxqueue->rps_flow_table);
		if (!flow_table)
			goto out;
		flow_id = skb->rxhash & flow_table->mask;
		rc = dev->netdev_ops->ndo_rx_flow_steer(dev, skb,
							rxq_index, flow_id);
		if (rc < 0)
			goto out;
		old_rflow = rflow;
		rflow = &flow_table->flows[flow_id];
		rflow->filter = rc;
		if (old_rflow->filter == rflow->filter)
			old_rflow->filter = RPS_NO_FILTER;
	out:
#endif
		rflow->last_qtail =
2879
			per_cpu(softnet_data, next_cpu).input_queue_head;
2880 2881
	}

2882
	rflow->cpu = next_cpu;
2883 2884 2885
	return rflow;
}

2886 2887 2888 2889 2890 2891 2892 2893 2894
/*
 * get_rps_cpu is called from netif_receive_skb and returns the target
 * CPU from the RPS map of the receiving queue for a given skb.
 * rcu_read_lock must be held on entry.
 */
static int get_rps_cpu(struct net_device *dev, struct sk_buff *skb,
		       struct rps_dev_flow **rflowp)
{
	struct netdev_rx_queue *rxqueue;
E
Eric Dumazet 已提交
2895
	struct rps_map *map;
2896 2897 2898 2899 2900 2901 2902
	struct rps_dev_flow_table *flow_table;
	struct rps_sock_flow_table *sock_flow_table;
	int cpu = -1;
	u16 tcpu;

	if (skb_rx_queue_recorded(skb)) {
		u16 index = skb_get_rx_queue(skb);
2903 2904 2905 2906 2907
		if (unlikely(index >= dev->real_num_rx_queues)) {
			WARN_ONCE(dev->real_num_rx_queues > 1,
				  "%s received packet on queue %u, but number "
				  "of RX queues is %u\n",
				  dev->name, index, dev->real_num_rx_queues);
2908 2909 2910 2911 2912 2913
			goto done;
		}
		rxqueue = dev->_rx + index;
	} else
		rxqueue = dev->_rx;

E
Eric Dumazet 已提交
2914 2915
	map = rcu_dereference(rxqueue->rps_map);
	if (map) {
2916
		if (map->len == 1 &&
2917
		    !rcu_access_pointer(rxqueue->rps_flow_table)) {
2918 2919 2920 2921 2922
			tcpu = map->cpus[0];
			if (cpu_online(tcpu))
				cpu = tcpu;
			goto done;
		}
2923
	} else if (!rcu_access_pointer(rxqueue->rps_flow_table)) {
2924
		goto done;
2925
	}
2926

2927
	skb_reset_network_header(skb);
2928 2929 2930
	if (!skb_get_rxhash(skb))
		goto done;

T
Tom Herbert 已提交
2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956
	flow_table = rcu_dereference(rxqueue->rps_flow_table);
	sock_flow_table = rcu_dereference(rps_sock_flow_table);
	if (flow_table && sock_flow_table) {
		u16 next_cpu;
		struct rps_dev_flow *rflow;

		rflow = &flow_table->flows[skb->rxhash & flow_table->mask];
		tcpu = rflow->cpu;

		next_cpu = sock_flow_table->ents[skb->rxhash &
		    sock_flow_table->mask];

		/*
		 * If the desired CPU (where last recvmsg was done) is
		 * different from current CPU (one in the rx-queue flow
		 * table entry), switch if one of the following holds:
		 *   - Current CPU is unset (equal to RPS_NO_CPU).
		 *   - Current CPU is offline.
		 *   - The current CPU's queue tail has advanced beyond the
		 *     last packet that was enqueued using this table entry.
		 *     This guarantees that all previous packets for the flow
		 *     have been dequeued, thus preserving in order delivery.
		 */
		if (unlikely(tcpu != next_cpu) &&
		    (tcpu == RPS_NO_CPU || !cpu_online(tcpu) ||
		     ((int)(per_cpu(softnet_data, tcpu).input_queue_head -
2957 2958
		      rflow->last_qtail)) >= 0)) {
			tcpu = next_cpu;
2959
			rflow = set_rps_cpu(dev, skb, rflow, next_cpu);
2960
		}
2961

T
Tom Herbert 已提交
2962 2963 2964 2965 2966 2967 2968
		if (tcpu != RPS_NO_CPU && cpu_online(tcpu)) {
			*rflowp = rflow;
			cpu = tcpu;
			goto done;
		}
	}

T
Tom Herbert 已提交
2969
	if (map) {
T
Tom Herbert 已提交
2970
		tcpu = map->cpus[((u64) skb->rxhash * map->len) >> 32];
T
Tom Herbert 已提交
2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981

		if (cpu_online(tcpu)) {
			cpu = tcpu;
			goto done;
		}
	}

done:
	return cpu;
}

2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021
#ifdef CONFIG_RFS_ACCEL

/**
 * rps_may_expire_flow - check whether an RFS hardware filter may be removed
 * @dev: Device on which the filter was set
 * @rxq_index: RX queue index
 * @flow_id: Flow ID passed to ndo_rx_flow_steer()
 * @filter_id: Filter ID returned by ndo_rx_flow_steer()
 *
 * Drivers that implement ndo_rx_flow_steer() should periodically call
 * this function for each installed filter and remove the filters for
 * which it returns %true.
 */
bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index,
			 u32 flow_id, u16 filter_id)
{
	struct netdev_rx_queue *rxqueue = dev->_rx + rxq_index;
	struct rps_dev_flow_table *flow_table;
	struct rps_dev_flow *rflow;
	bool expire = true;
	int cpu;

	rcu_read_lock();
	flow_table = rcu_dereference(rxqueue->rps_flow_table);
	if (flow_table && flow_id <= flow_table->mask) {
		rflow = &flow_table->flows[flow_id];
		cpu = ACCESS_ONCE(rflow->cpu);
		if (rflow->filter == filter_id && cpu != RPS_NO_CPU &&
		    ((int)(per_cpu(softnet_data, cpu).input_queue_head -
			   rflow->last_qtail) <
		     (int)(10 * flow_table->mask)))
			expire = false;
	}
	rcu_read_unlock();
	return expire;
}
EXPORT_SYMBOL(rps_may_expire_flow);

#endif /* CONFIG_RFS_ACCEL */

T
Tom Herbert 已提交
3022
/* Called from hardirq (IPI) context */
E
Eric Dumazet 已提交
3023
static void rps_trigger_softirq(void *data)
T
Tom Herbert 已提交
3024
{
E
Eric Dumazet 已提交
3025 3026
	struct softnet_data *sd = data;

E
Eric Dumazet 已提交
3027
	____napi_schedule(sd, &sd->backlog);
C
Changli Gao 已提交
3028
	sd->received_rps++;
T
Tom Herbert 已提交
3029
}
E
Eric Dumazet 已提交
3030

T
Tom Herbert 已提交
3031
#endif /* CONFIG_RPS */
T
Tom Herbert 已提交
3032

E
Eric Dumazet 已提交
3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053
/*
 * Check if this softnet_data structure is another cpu one
 * If yes, queue it to our IPI list and return 1
 * If no, return 0
 */
static int rps_ipi_queued(struct softnet_data *sd)
{
#ifdef CONFIG_RPS
	struct softnet_data *mysd = &__get_cpu_var(softnet_data);

	if (sd != mysd) {
		sd->rps_ipi_next = mysd->rps_ipi_list;
		mysd->rps_ipi_list = sd;

		__raise_softirq_irqoff(NET_RX_SOFTIRQ);
		return 1;
	}
#endif /* CONFIG_RPS */
	return 0;
}

T
Tom Herbert 已提交
3054 3055 3056 3057
/*
 * enqueue_to_backlog is called to queue an skb to a per CPU backlog
 * queue (may be a remote CPU queue).
 */
T
Tom Herbert 已提交
3058 3059
static int enqueue_to_backlog(struct sk_buff *skb, int cpu,
			      unsigned int *qtail)
T
Tom Herbert 已提交
3060
{
E
Eric Dumazet 已提交
3061
	struct softnet_data *sd;
T
Tom Herbert 已提交
3062 3063
	unsigned long flags;

E
Eric Dumazet 已提交
3064
	sd = &per_cpu(softnet_data, cpu);
T
Tom Herbert 已提交
3065 3066 3067

	local_irq_save(flags);

E
Eric Dumazet 已提交
3068
	rps_lock(sd);
3069 3070
	if (skb_queue_len(&sd->input_pkt_queue) <= netdev_max_backlog) {
		if (skb_queue_len(&sd->input_pkt_queue)) {
T
Tom Herbert 已提交
3071
enqueue:
E
Eric Dumazet 已提交
3072
			__skb_queue_tail(&sd->input_pkt_queue, skb);
3073
			input_queue_tail_incr_save(sd, qtail);
E
Eric Dumazet 已提交
3074
			rps_unlock(sd);
3075
			local_irq_restore(flags);
T
Tom Herbert 已提交
3076 3077 3078
			return NET_RX_SUCCESS;
		}

3079 3080 3081 3082
		/* Schedule NAPI for backlog device
		 * We can use non atomic operation since we own the queue lock
		 */
		if (!__test_and_set_bit(NAPI_STATE_SCHED, &sd->backlog.state)) {
E
Eric Dumazet 已提交
3083
			if (!rps_ipi_queued(sd))
E
Eric Dumazet 已提交
3084
				____napi_schedule(sd, &sd->backlog);
T
Tom Herbert 已提交
3085 3086 3087 3088
		}
		goto enqueue;
	}

C
Changli Gao 已提交
3089
	sd->dropped++;
E
Eric Dumazet 已提交
3090
	rps_unlock(sd);
T
Tom Herbert 已提交
3091 3092 3093

	local_irq_restore(flags);

3094
	atomic_long_inc(&skb->dev->rx_dropped);
T
Tom Herbert 已提交
3095 3096 3097
	kfree_skb(skb);
	return NET_RX_DROP;
}
L
Linus Torvalds 已提交
3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115

/**
 *	netif_rx	-	post buffer to the network code
 *	@skb: buffer to post
 *
 *	This function receives a packet from a device driver and queues it for
 *	the upper (protocol) levels to process.  It always succeeds. The buffer
 *	may be dropped during processing for congestion control or by the
 *	protocol layers.
 *
 *	return values:
 *	NET_RX_SUCCESS	(no congestion)
 *	NET_RX_DROP     (packet was dropped)
 *
 */

int netif_rx(struct sk_buff *skb)
{
3116
	int ret;
L
Linus Torvalds 已提交
3117 3118 3119 3120 3121

	/* if netpoll wants it, pretend we never saw it */
	if (netpoll_rx(skb))
		return NET_RX_DROP;

3122
	net_timestamp_check(netdev_tstamp_prequeue, skb);
L
Linus Torvalds 已提交
3123

3124
	trace_netif_rx(skb);
E
Eric Dumazet 已提交
3125
#ifdef CONFIG_RPS
3126
	if (static_key_false(&rps_needed)) {
T
Tom Herbert 已提交
3127
		struct rps_dev_flow voidflow, *rflow = &voidflow;
3128 3129
		int cpu;

3130
		preempt_disable();
3131
		rcu_read_lock();
T
Tom Herbert 已提交
3132 3133

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
3134 3135
		if (cpu < 0)
			cpu = smp_processor_id();
T
Tom Herbert 已提交
3136 3137 3138

		ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);

3139
		rcu_read_unlock();
3140
		preempt_enable();
3141 3142
	} else
#endif
T
Tom Herbert 已提交
3143 3144 3145 3146 3147
	{
		unsigned int qtail;
		ret = enqueue_to_backlog(skb, get_cpu(), &qtail);
		put_cpu();
	}
3148
	return ret;
L
Linus Torvalds 已提交
3149
}
E
Eric Dumazet 已提交
3150
EXPORT_SYMBOL(netif_rx);
L
Linus Torvalds 已提交
3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181

int netif_rx_ni(struct sk_buff *skb)
{
	int err;

	preempt_disable();
	err = netif_rx(skb);
	if (local_softirq_pending())
		do_softirq();
	preempt_enable();

	return err;
}
EXPORT_SYMBOL(netif_rx_ni);

static void net_tx_action(struct softirq_action *h)
{
	struct softnet_data *sd = &__get_cpu_var(softnet_data);

	if (sd->completion_queue) {
		struct sk_buff *clist;

		local_irq_disable();
		clist = sd->completion_queue;
		sd->completion_queue = NULL;
		local_irq_enable();

		while (clist) {
			struct sk_buff *skb = clist;
			clist = clist->next;

3182
			WARN_ON(atomic_read(&skb->users));
3183
			trace_kfree_skb(skb, net_tx_action);
L
Linus Torvalds 已提交
3184 3185 3186 3187 3188
			__kfree_skb(skb);
		}
	}

	if (sd->output_queue) {
3189
		struct Qdisc *head;
L
Linus Torvalds 已提交
3190 3191 3192 3193

		local_irq_disable();
		head = sd->output_queue;
		sd->output_queue = NULL;
3194
		sd->output_queue_tailp = &sd->output_queue;
L
Linus Torvalds 已提交
3195 3196 3197
		local_irq_enable();

		while (head) {
3198 3199 3200
			struct Qdisc *q = head;
			spinlock_t *root_lock;

L
Linus Torvalds 已提交
3201 3202
			head = head->next_sched;

3203
			root_lock = qdisc_lock(q);
3204
			if (spin_trylock(root_lock)) {
3205 3206 3207
				smp_mb__before_clear_bit();
				clear_bit(__QDISC_STATE_SCHED,
					  &q->state);
3208 3209
				qdisc_run(q);
				spin_unlock(root_lock);
L
Linus Torvalds 已提交
3210
			} else {
3211
				if (!test_bit(__QDISC_STATE_DEACTIVATED,
3212
					      &q->state)) {
3213
					__netif_reschedule(q);
3214 3215 3216 3217 3218
				} else {
					smp_mb__before_clear_bit();
					clear_bit(__QDISC_STATE_SCHED,
						  &q->state);
				}
L
Linus Torvalds 已提交
3219 3220 3221 3222 3223
			}
		}
	}
}

3224 3225
#if (defined(CONFIG_BRIDGE) || defined(CONFIG_BRIDGE_MODULE)) && \
    (defined(CONFIG_ATM_LANE) || defined(CONFIG_ATM_LANE_MODULE))
3226 3227 3228
/* This hook is defined here for ATM LANE */
int (*br_fdb_test_addr_hook)(struct net_device *dev,
			     unsigned char *addr) __read_mostly;
3229
EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook);
3230
#endif
L
Linus Torvalds 已提交
3231 3232 3233 3234 3235 3236

#ifdef CONFIG_NET_CLS_ACT
/* TODO: Maybe we should just force sch_ingress to be compiled in
 * when CONFIG_NET_CLS_ACT is? otherwise some useless instructions
 * a compare and 2 stores extra right now if we dont have it on
 * but have CONFIG_NET_CLS_ACT
L
Lucas De Marchi 已提交
3237 3238
 * NOTE: This doesn't stop any functionality; if you dont have
 * the ingress scheduler, you just can't add policies on ingress.
L
Linus Torvalds 已提交
3239 3240
 *
 */
3241
static int ing_filter(struct sk_buff *skb, struct netdev_queue *rxq)
L
Linus Torvalds 已提交
3242 3243
{
	struct net_device *dev = skb->dev;
3244
	u32 ttl = G_TC_RTTL(skb->tc_verd);
3245 3246
	int result = TC_ACT_OK;
	struct Qdisc *q;
3247

3248
	if (unlikely(MAX_RED_LOOP < ttl++)) {
3249 3250
		net_warn_ratelimited("Redir loop detected Dropping packet (%d->%d)\n",
				     skb->skb_iif, dev->ifindex);
3251 3252
		return TC_ACT_SHOT;
	}
L
Linus Torvalds 已提交
3253

3254 3255
	skb->tc_verd = SET_TC_RTTL(skb->tc_verd, ttl);
	skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_INGRESS);
L
Linus Torvalds 已提交
3256

3257
	q = rxq->qdisc;
3258
	if (q != &noop_qdisc) {
3259
		spin_lock(qdisc_lock(q));
3260 3261
		if (likely(!test_bit(__QDISC_STATE_DEACTIVATED, &q->state)))
			result = qdisc_enqueue_root(skb, q);
3262 3263
		spin_unlock(qdisc_lock(q));
	}
3264 3265 3266

	return result;
}
3267

3268 3269 3270 3271
static inline struct sk_buff *handle_ing(struct sk_buff *skb,
					 struct packet_type **pt_prev,
					 int *ret, struct net_device *orig_dev)
{
3272 3273 3274
	struct netdev_queue *rxq = rcu_dereference(skb->dev->ingress_queue);

	if (!rxq || rxq->qdisc == &noop_qdisc)
3275
		goto out;
L
Linus Torvalds 已提交
3276

3277 3278 3279
	if (*pt_prev) {
		*ret = deliver_skb(skb, *pt_prev, orig_dev);
		*pt_prev = NULL;
L
Linus Torvalds 已提交
3280 3281
	}

3282
	switch (ing_filter(skb, rxq)) {
3283 3284 3285 3286 3287 3288 3289 3290 3291
	case TC_ACT_SHOT:
	case TC_ACT_STOLEN:
		kfree_skb(skb);
		return NULL;
	}

out:
	skb->tc_verd = 0;
	return skb;
L
Linus Torvalds 已提交
3292 3293 3294
}
#endif

3295 3296 3297 3298
/**
 *	netdev_rx_handler_register - register receive handler
 *	@dev: device to register a handler for
 *	@rx_handler: receive handler to register
J
Jiri Pirko 已提交
3299
 *	@rx_handler_data: data pointer that is used by rx handler
3300 3301 3302 3303 3304 3305
 *
 *	Register a receive hander for a device. This handler will then be
 *	called from __netif_receive_skb. A negative errno code is returned
 *	on a failure.
 *
 *	The caller must hold the rtnl_mutex.
3306 3307
 *
 *	For a general description of rx_handler, see enum rx_handler_result.
3308 3309
 */
int netdev_rx_handler_register(struct net_device *dev,
J
Jiri Pirko 已提交
3310 3311
			       rx_handler_func_t *rx_handler,
			       void *rx_handler_data)
3312 3313 3314 3315 3316 3317
{
	ASSERT_RTNL();

	if (dev->rx_handler)
		return -EBUSY;

J
Jiri Pirko 已提交
3318
	rcu_assign_pointer(dev->rx_handler_data, rx_handler_data);
3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336
	rcu_assign_pointer(dev->rx_handler, rx_handler);

	return 0;
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_register);

/**
 *	netdev_rx_handler_unregister - unregister receive handler
 *	@dev: device to unregister a handler from
 *
 *	Unregister a receive hander from a device.
 *
 *	The caller must hold the rtnl_mutex.
 */
void netdev_rx_handler_unregister(struct net_device *dev)
{

	ASSERT_RTNL();
3337 3338
	RCU_INIT_POINTER(dev->rx_handler, NULL);
	RCU_INIT_POINTER(dev->rx_handler_data, NULL);
3339 3340 3341
}
EXPORT_SYMBOL_GPL(netdev_rx_handler_unregister);

3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358
/*
 * Limit the use of PFMEMALLOC reserves to those protocols that implement
 * the special handling of PFMEMALLOC skbs.
 */
static bool skb_pfmemalloc_protocol(struct sk_buff *skb)
{
	switch (skb->protocol) {
	case __constant_htons(ETH_P_ARP):
	case __constant_htons(ETH_P_IP):
	case __constant_htons(ETH_P_IPV6):
	case __constant_htons(ETH_P_8021Q):
		return true;
	default:
		return false;
	}
}

3359
static int __netif_receive_skb_core(struct sk_buff *skb, bool pfmemalloc)
L
Linus Torvalds 已提交
3360 3361
{
	struct packet_type *ptype, *pt_prev;
3362
	rx_handler_func_t *rx_handler;
D
David S. Miller 已提交
3363
	struct net_device *orig_dev;
3364
	struct net_device *null_or_dev;
3365
	bool deliver_exact = false;
L
Linus Torvalds 已提交
3366
	int ret = NET_RX_DROP;
A
Al Viro 已提交
3367
	__be16 type;
L
Linus Torvalds 已提交
3368

3369
	net_timestamp_check(!netdev_tstamp_prequeue, skb);
3370

3371
	trace_netif_receive_skb(skb);
3372

L
Linus Torvalds 已提交
3373
	/* if we've gotten here through NAPI, check netpoll */
3374
	if (netpoll_receive_skb(skb))
3375
		goto out;
L
Linus Torvalds 已提交
3376

J
Joe Eykholt 已提交
3377
	orig_dev = skb->dev;
3378

3379
	skb_reset_network_header(skb);
3380 3381
	if (!skb_transport_header_was_set(skb))
		skb_reset_transport_header(skb);
3382
	skb_reset_mac_len(skb);
L
Linus Torvalds 已提交
3383 3384 3385 3386 3387

	pt_prev = NULL;

	rcu_read_lock();

3388
another_round:
3389
	skb->skb_iif = skb->dev->ifindex;
3390 3391 3392

	__this_cpu_inc(softnet_data.processed);

3393 3394 3395
	if (skb->protocol == cpu_to_be16(ETH_P_8021Q)) {
		skb = vlan_untag(skb);
		if (unlikely(!skb))
3396
			goto unlock;
3397 3398
	}

L
Linus Torvalds 已提交
3399 3400 3401 3402 3403 3404 3405
#ifdef CONFIG_NET_CLS_ACT
	if (skb->tc_verd & TC_NCLS) {
		skb->tc_verd = CLR_TC_NCLS(skb->tc_verd);
		goto ncls;
	}
#endif

3406
	if (pfmemalloc)
3407 3408
		goto skip_taps;

L
Linus Torvalds 已提交
3409
	list_for_each_entry_rcu(ptype, &ptype_all, list) {
3410
		if (!ptype->dev || ptype->dev == skb->dev) {
3411
			if (pt_prev)
D
David S. Miller 已提交
3412
				ret = deliver_skb(skb, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3413 3414 3415 3416
			pt_prev = ptype;
		}
	}

3417
skip_taps:
L
Linus Torvalds 已提交
3418
#ifdef CONFIG_NET_CLS_ACT
3419 3420
	skb = handle_ing(skb, &pt_prev, &ret, orig_dev);
	if (!skb)
3421
		goto unlock;
L
Linus Torvalds 已提交
3422 3423 3424
ncls:
#endif

3425
	if (pfmemalloc && !skb_pfmemalloc_protocol(skb))
3426 3427
		goto drop;

3428 3429 3430 3431 3432
	if (vlan_tx_tag_present(skb)) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3433
		if (vlan_do_receive(&skb))
3434 3435
			goto another_round;
		else if (unlikely(!skb))
3436
			goto unlock;
3437 3438
	}

3439
	rx_handler = rcu_dereference(skb->dev->rx_handler);
3440 3441 3442 3443 3444
	if (rx_handler) {
		if (pt_prev) {
			ret = deliver_skb(skb, pt_prev, orig_dev);
			pt_prev = NULL;
		}
3445 3446
		switch (rx_handler(&skb)) {
		case RX_HANDLER_CONSUMED:
3447
			goto unlock;
3448
		case RX_HANDLER_ANOTHER:
3449
			goto another_round;
3450 3451 3452 3453 3454 3455 3456
		case RX_HANDLER_EXACT:
			deliver_exact = true;
		case RX_HANDLER_PASS:
			break;
		default:
			BUG();
		}
3457
	}
L
Linus Torvalds 已提交
3458

3459 3460 3461
	if (vlan_tx_nonzero_tag_present(skb))
		skb->pkt_type = PACKET_OTHERHOST;

3462
	/* deliver only exact match when indicated */
3463
	null_or_dev = deliver_exact ? skb->dev : NULL;
3464

L
Linus Torvalds 已提交
3465
	type = skb->protocol;
3466 3467
	list_for_each_entry_rcu(ptype,
			&ptype_base[ntohs(type) & PTYPE_HASH_MASK], list) {
3468
		if (ptype->type == type &&
3469 3470
		    (ptype->dev == null_or_dev || ptype->dev == skb->dev ||
		     ptype->dev == orig_dev)) {
3471
			if (pt_prev)
D
David S. Miller 已提交
3472
				ret = deliver_skb(skb, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3473 3474 3475 3476 3477
			pt_prev = ptype;
		}
	}

	if (pt_prev) {
3478
		if (unlikely(skb_orphan_frags(skb, GFP_ATOMIC)))
3479
			goto drop;
3480 3481
		else
			ret = pt_prev->func(skb, skb->dev, pt_prev, orig_dev);
L
Linus Torvalds 已提交
3482
	} else {
3483
drop:
3484
		atomic_long_inc(&skb->dev->rx_dropped);
L
Linus Torvalds 已提交
3485 3486 3487 3488 3489 3490 3491
		kfree_skb(skb);
		/* Jamal, now you will not able to escape explaining
		 * me how you were going to use this. :-)
		 */
		ret = NET_RX_DROP;
	}

3492
unlock:
L
Linus Torvalds 已提交
3493
	rcu_read_unlock();
3494
out:
3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519
	return ret;
}

static int __netif_receive_skb(struct sk_buff *skb)
{
	int ret;

	if (sk_memalloc_socks() && skb_pfmemalloc(skb)) {
		unsigned long pflags = current->flags;

		/*
		 * PFMEMALLOC skbs are special, they should
		 * - be delivered to SOCK_MEMALLOC sockets only
		 * - stay away from userspace
		 * - have bounded memory usage
		 *
		 * Use PF_MEMALLOC as this saves us from propagating the allocation
		 * context down to all allocation sites.
		 */
		current->flags |= PF_MEMALLOC;
		ret = __netif_receive_skb_core(skb, true);
		tsk_restore_flags(current, pflags, PF_MEMALLOC);
	} else
		ret = __netif_receive_skb_core(skb, false);

L
Linus Torvalds 已提交
3520 3521
	return ret;
}
T
Tom Herbert 已提交
3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539

/**
 *	netif_receive_skb - process receive buffer from network
 *	@skb: buffer to process
 *
 *	netif_receive_skb() is the main receive data processing function.
 *	It always succeeds. The buffer may be dropped during processing
 *	for congestion control or by the protocol layers.
 *
 *	This function may only be called from softirq context and interrupts
 *	should be enabled.
 *
 *	Return values (usually ignored):
 *	NET_RX_SUCCESS: no congestion
 *	NET_RX_DROP: packet was dropped
 */
int netif_receive_skb(struct sk_buff *skb)
{
3540
	net_timestamp_check(netdev_tstamp_prequeue, skb);
E
Eric Dumazet 已提交
3541

3542 3543 3544
	if (skb_defer_rx_timestamp(skb))
		return NET_RX_SUCCESS;

E
Eric Dumazet 已提交
3545
#ifdef CONFIG_RPS
3546
	if (static_key_false(&rps_needed)) {
E
Eric Dumazet 已提交
3547 3548
		struct rps_dev_flow voidflow, *rflow = &voidflow;
		int cpu, ret;
T
Tom Herbert 已提交
3549

E
Eric Dumazet 已提交
3550 3551 3552
		rcu_read_lock();

		cpu = get_rps_cpu(skb->dev, skb, &rflow);
T
Tom Herbert 已提交
3553

E
Eric Dumazet 已提交
3554 3555 3556
		if (cpu >= 0) {
			ret = enqueue_to_backlog(skb, cpu, &rflow->last_qtail);
			rcu_read_unlock();
3557
			return ret;
E
Eric Dumazet 已提交
3558
		}
3559
		rcu_read_unlock();
T
Tom Herbert 已提交
3560
	}
3561
#endif
3562
	return __netif_receive_skb(skb);
T
Tom Herbert 已提交
3563
}
E
Eric Dumazet 已提交
3564
EXPORT_SYMBOL(netif_receive_skb);
L
Linus Torvalds 已提交
3565

E
Eric Dumazet 已提交
3566 3567 3568
/* Network device is going away, flush any packets still pending
 * Called with irqs disabled.
 */
3569
static void flush_backlog(void *arg)
3570
{
3571
	struct net_device *dev = arg;
E
Eric Dumazet 已提交
3572
	struct softnet_data *sd = &__get_cpu_var(softnet_data);
3573 3574
	struct sk_buff *skb, *tmp;

E
Eric Dumazet 已提交
3575
	rps_lock(sd);
3576
	skb_queue_walk_safe(&sd->input_pkt_queue, skb, tmp) {
3577
		if (skb->dev == dev) {
E
Eric Dumazet 已提交
3578
			__skb_unlink(skb, &sd->input_pkt_queue);
3579
			kfree_skb(skb);
3580
			input_queue_head_incr(sd);
3581
		}
3582
	}
E
Eric Dumazet 已提交
3583
	rps_unlock(sd);
3584 3585 3586 3587 3588

	skb_queue_walk_safe(&sd->process_queue, skb, tmp) {
		if (skb->dev == dev) {
			__skb_unlink(skb, &sd->process_queue);
			kfree_skb(skb);
3589
			input_queue_head_incr(sd);
3590 3591
		}
	}
3592 3593
}

3594 3595
static int napi_gro_complete(struct sk_buff *skb)
{
3596
	struct packet_offload *ptype;
3597
	__be16 type = skb->protocol;
3598
	struct list_head *head = &offload_base;
3599 3600
	int err = -ENOENT;

3601 3602
	BUILD_BUG_ON(sizeof(struct napi_gro_cb) > sizeof(skb->cb));

3603 3604
	if (NAPI_GRO_CB(skb)->count == 1) {
		skb_shinfo(skb)->gso_size = 0;
3605
		goto out;
3606
	}
3607 3608 3609

	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
3610
		if (ptype->type != type || !ptype->callbacks.gro_complete)
3611 3612
			continue;

3613
		err = ptype->callbacks.gro_complete(skb);
3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627
		break;
	}
	rcu_read_unlock();

	if (err) {
		WARN_ON(&ptype->list == head);
		kfree_skb(skb);
		return NET_RX_SUCCESS;
	}

out:
	return netif_receive_skb(skb);
}

3628 3629 3630 3631 3632
/* napi->gro_list contains packets ordered by age.
 * youngest packets at the head of it.
 * Complete skbs in reverse order to reduce latencies.
 */
void napi_gro_flush(struct napi_struct *napi, bool flush_old)
3633
{
3634
	struct sk_buff *skb, *prev = NULL;
3635

3636 3637 3638 3639 3640 3641 3642
	/* scan list and build reverse chain */
	for (skb = napi->gro_list; skb != NULL; skb = skb->next) {
		skb->prev = prev;
		prev = skb;
	}

	for (skb = prev; skb; skb = prev) {
3643
		skb->next = NULL;
3644 3645 3646 3647 3648

		if (flush_old && NAPI_GRO_CB(skb)->age == jiffies)
			return;

		prev = skb->prev;
3649
		napi_gro_complete(skb);
3650
		napi->gro_count--;
3651 3652 3653 3654
	}

	napi->gro_list = NULL;
}
E
Eric Dumazet 已提交
3655
EXPORT_SYMBOL(napi_gro_flush);
3656

E
Eric Dumazet 已提交
3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678
static void gro_list_prepare(struct napi_struct *napi, struct sk_buff *skb)
{
	struct sk_buff *p;
	unsigned int maclen = skb->dev->hard_header_len;

	for (p = napi->gro_list; p; p = p->next) {
		unsigned long diffs;

		diffs = (unsigned long)p->dev ^ (unsigned long)skb->dev;
		diffs |= p->vlan_tci ^ skb->vlan_tci;
		if (maclen == ETH_HLEN)
			diffs |= compare_ether_header(skb_mac_header(p),
						      skb_gro_mac_header(skb));
		else if (!diffs)
			diffs = memcmp(skb_mac_header(p),
				       skb_gro_mac_header(skb),
				       maclen);
		NAPI_GRO_CB(p)->same_flow = !diffs;
		NAPI_GRO_CB(p)->flush = 0;
	}
}

R
Rami Rosen 已提交
3679
static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
3680 3681
{
	struct sk_buff **pp = NULL;
3682
	struct packet_offload *ptype;
3683
	__be16 type = skb->protocol;
3684
	struct list_head *head = &offload_base;
H
Herbert Xu 已提交
3685
	int same_flow;
3686
	enum gro_result ret;
3687

3688
	if (!(skb->dev->features & NETIF_F_GRO) || netpoll_rx_on(skb))
3689 3690
		goto normal;

3691
	if (skb_is_gso(skb) || skb_has_frag_list(skb))
3692 3693
		goto normal;

E
Eric Dumazet 已提交
3694 3695
	gro_list_prepare(napi, skb);

3696 3697
	rcu_read_lock();
	list_for_each_entry_rcu(ptype, head, list) {
3698
		if (ptype->type != type || !ptype->callbacks.gro_receive)
3699 3700
			continue;

3701
		skb_set_network_header(skb, skb_gro_offset(skb));
3702
		skb_reset_mac_len(skb);
3703 3704
		NAPI_GRO_CB(skb)->same_flow = 0;
		NAPI_GRO_CB(skb)->flush = 0;
H
Herbert Xu 已提交
3705
		NAPI_GRO_CB(skb)->free = 0;
3706

3707
		pp = ptype->callbacks.gro_receive(&napi->gro_list, skb);
3708 3709 3710 3711 3712 3713 3714
		break;
	}
	rcu_read_unlock();

	if (&ptype->list == head)
		goto normal;

H
Herbert Xu 已提交
3715
	same_flow = NAPI_GRO_CB(skb)->same_flow;
3716
	ret = NAPI_GRO_CB(skb)->free ? GRO_MERGED_FREE : GRO_MERGED;
H
Herbert Xu 已提交
3717

3718 3719 3720 3721 3722 3723
	if (pp) {
		struct sk_buff *nskb = *pp;

		*pp = nskb->next;
		nskb->next = NULL;
		napi_gro_complete(nskb);
3724
		napi->gro_count--;
3725 3726
	}

H
Herbert Xu 已提交
3727
	if (same_flow)
3728 3729
		goto ok;

3730
	if (NAPI_GRO_CB(skb)->flush || napi->gro_count >= MAX_GRO_SKBS)
3731 3732
		goto normal;

3733
	napi->gro_count++;
3734
	NAPI_GRO_CB(skb)->count = 1;
3735
	NAPI_GRO_CB(skb)->age = jiffies;
3736
	skb_shinfo(skb)->gso_size = skb_gro_len(skb);
3737 3738
	skb->next = napi->gro_list;
	napi->gro_list = skb;
3739
	ret = GRO_HELD;
3740

3741
pull:
H
Herbert Xu 已提交
3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752
	if (skb_headlen(skb) < skb_gro_offset(skb)) {
		int grow = skb_gro_offset(skb) - skb_headlen(skb);

		BUG_ON(skb->end - skb->tail < grow);

		memcpy(skb_tail_pointer(skb), NAPI_GRO_CB(skb)->frag0, grow);

		skb->tail += grow;
		skb->data_len -= grow;

		skb_shinfo(skb)->frags[0].page_offset += grow;
E
Eric Dumazet 已提交
3753
		skb_frag_size_sub(&skb_shinfo(skb)->frags[0], grow);
H
Herbert Xu 已提交
3754

E
Eric Dumazet 已提交
3755
		if (unlikely(!skb_frag_size(&skb_shinfo(skb)->frags[0]))) {
3756
			skb_frag_unref(skb, 0);
H
Herbert Xu 已提交
3757 3758
			memmove(skb_shinfo(skb)->frags,
				skb_shinfo(skb)->frags + 1,
3759
				--skb_shinfo(skb)->nr_frags * sizeof(skb_frag_t));
H
Herbert Xu 已提交
3760
		}
3761 3762
	}

3763
ok:
3764
	return ret;
3765 3766

normal:
3767 3768
	ret = GRO_NORMAL;
	goto pull;
H
Herbert Xu 已提交
3769
}
3770

H
Herbert Xu 已提交
3771

R
Rami Rosen 已提交
3772
static gro_result_t napi_skb_finish(gro_result_t ret, struct sk_buff *skb)
H
Herbert Xu 已提交
3773
{
3774 3775
	switch (ret) {
	case GRO_NORMAL:
3776 3777 3778
		if (netif_receive_skb(skb))
			ret = GRO_DROP;
		break;
H
Herbert Xu 已提交
3779

3780
	case GRO_DROP:
H
Herbert Xu 已提交
3781 3782
		kfree_skb(skb);
		break;
3783

3784
	case GRO_MERGED_FREE:
3785 3786 3787 3788
		if (NAPI_GRO_CB(skb)->free == NAPI_GRO_FREE_STOLEN_HEAD)
			kmem_cache_free(skbuff_head_cache, skb);
		else
			__kfree_skb(skb);
3789 3790
		break;

3791 3792 3793
	case GRO_HELD:
	case GRO_MERGED:
		break;
H
Herbert Xu 已提交
3794 3795
	}

3796
	return ret;
3797 3798
}

3799
static void skb_gro_reset_offset(struct sk_buff *skb)
3800
{
3801 3802 3803
	const struct skb_shared_info *pinfo = skb_shinfo(skb);
	const skb_frag_t *frag0 = &pinfo->frags[0];

3804 3805
	NAPI_GRO_CB(skb)->data_offset = 0;
	NAPI_GRO_CB(skb)->frag0 = NULL;
3806
	NAPI_GRO_CB(skb)->frag0_len = 0;
3807

3808
	if (skb->mac_header == skb->tail &&
3809 3810 3811 3812
	    pinfo->nr_frags &&
	    !PageHighMem(skb_frag_page(frag0))) {
		NAPI_GRO_CB(skb)->frag0 = skb_frag_address(frag0);
		NAPI_GRO_CB(skb)->frag0_len = skb_frag_size(frag0);
3813
	}
3814 3815
}

3816
gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb)
3817
{
3818 3819
	skb_gro_reset_offset(skb);

E
Eric Dumazet 已提交
3820
	return napi_skb_finish(dev_gro_receive(napi, skb), skb);
3821 3822 3823
}
EXPORT_SYMBOL(napi_gro_receive);

S
stephen hemminger 已提交
3824
static void napi_reuse_skb(struct napi_struct *napi, struct sk_buff *skb)
3825 3826
{
	__skb_pull(skb, skb_headlen(skb));
3827 3828
	/* restore the reserve we had after netdev_alloc_skb_ip_align() */
	skb_reserve(skb, NET_SKB_PAD + NET_IP_ALIGN - skb_headroom(skb));
3829
	skb->vlan_tci = 0;
H
Herbert Xu 已提交
3830
	skb->dev = napi->dev;
A
Andy Gospodarek 已提交
3831
	skb->skb_iif = 0;
3832 3833 3834 3835

	napi->skb = skb;
}

3836
struct sk_buff *napi_get_frags(struct napi_struct *napi)
H
Herbert Xu 已提交
3837 3838 3839 3840
{
	struct sk_buff *skb = napi->skb;

	if (!skb) {
3841 3842 3843
		skb = netdev_alloc_skb_ip_align(napi->dev, GRO_MAX_HEAD);
		if (skb)
			napi->skb = skb;
3844
	}
3845 3846
	return skb;
}
3847
EXPORT_SYMBOL(napi_get_frags);
3848

R
Rami Rosen 已提交
3849
static gro_result_t napi_frags_finish(struct napi_struct *napi, struct sk_buff *skb,
3850
			       gro_result_t ret)
3851
{
3852 3853
	switch (ret) {
	case GRO_NORMAL:
3854
	case GRO_HELD:
3855
		skb->protocol = eth_type_trans(skb, skb->dev);
3856

3857 3858 3859 3860
		if (ret == GRO_HELD)
			skb_gro_pull(skb, -ETH_HLEN);
		else if (netif_receive_skb(skb))
			ret = GRO_DROP;
3861
		break;
H
Herbert Xu 已提交
3862

3863 3864 3865 3866
	case GRO_DROP:
	case GRO_MERGED_FREE:
		napi_reuse_skb(napi, skb);
		break;
3867 3868 3869

	case GRO_MERGED:
		break;
3870
	}
H
Herbert Xu 已提交
3871

3872
	return ret;
H
Herbert Xu 已提交
3873
}
3874

E
Eric Dumazet 已提交
3875
static struct sk_buff *napi_frags_skb(struct napi_struct *napi)
3876 3877 3878
{
	struct sk_buff *skb = napi->skb;
	struct ethhdr *eth;
3879 3880
	unsigned int hlen;
	unsigned int off;
3881 3882 3883 3884 3885 3886

	napi->skb = NULL;

	skb_reset_mac_header(skb);
	skb_gro_reset_offset(skb);

3887 3888 3889 3890 3891 3892 3893 3894 3895 3896
	off = skb_gro_offset(skb);
	hlen = off + sizeof(*eth);
	eth = skb_gro_header_fast(skb, off);
	if (skb_gro_header_hard(skb, hlen)) {
		eth = skb_gro_header_slow(skb, hlen, off);
		if (unlikely(!eth)) {
			napi_reuse_skb(napi, skb);
			skb = NULL;
			goto out;
		}
3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910
	}

	skb_gro_pull(skb, sizeof(*eth));

	/*
	 * This works because the only protocols we care about don't require
	 * special handling.  We'll fix it up properly at the end.
	 */
	skb->protocol = eth->h_proto;

out:
	return skb;
}

3911
gro_result_t napi_gro_frags(struct napi_struct *napi)
3912
{
3913
	struct sk_buff *skb = napi_frags_skb(napi);
3914 3915

	if (!skb)
3916
		return GRO_DROP;
3917

E
Eric Dumazet 已提交
3918
	return napi_frags_finish(napi, skb, dev_gro_receive(napi, skb));
3919
}
H
Herbert Xu 已提交
3920 3921
EXPORT_SYMBOL(napi_gro_frags);

3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949
/*
 * net_rps_action sends any pending IPI's for rps.
 * Note: called with local irq disabled, but exits with local irq enabled.
 */
static void net_rps_action_and_irq_enable(struct softnet_data *sd)
{
#ifdef CONFIG_RPS
	struct softnet_data *remsd = sd->rps_ipi_list;

	if (remsd) {
		sd->rps_ipi_list = NULL;

		local_irq_enable();

		/* Send pending IPI's to kick RPS processing on remote cpus. */
		while (remsd) {
			struct softnet_data *next = remsd->rps_ipi_next;

			if (cpu_online(remsd->cpu))
				__smp_call_function_single(remsd->cpu,
							   &remsd->csd, 0);
			remsd = next;
		}
	} else
#endif
		local_irq_enable();
}

3950
static int process_backlog(struct napi_struct *napi, int quota)
L
Linus Torvalds 已提交
3951 3952
{
	int work = 0;
E
Eric Dumazet 已提交
3953
	struct softnet_data *sd = container_of(napi, struct softnet_data, backlog);
L
Linus Torvalds 已提交
3954

3955 3956 3957 3958 3959 3960 3961 3962 3963
#ifdef CONFIG_RPS
	/* Check if we have pending ipi, its better to send them now,
	 * not waiting net_rx_action() end.
	 */
	if (sd->rps_ipi_list) {
		local_irq_disable();
		net_rps_action_and_irq_enable(sd);
	}
#endif
3964
	napi->weight = weight_p;
3965 3966
	local_irq_disable();
	while (work < quota) {
L
Linus Torvalds 已提交
3967
		struct sk_buff *skb;
3968 3969 3970 3971 3972 3973
		unsigned int qlen;

		while ((skb = __skb_dequeue(&sd->process_queue))) {
			local_irq_enable();
			__netif_receive_skb(skb);
			local_irq_disable();
3974 3975 3976 3977 3978
			input_queue_head_incr(sd);
			if (++work >= quota) {
				local_irq_enable();
				return work;
			}
3979
		}
L
Linus Torvalds 已提交
3980

E
Eric Dumazet 已提交
3981
		rps_lock(sd);
3982
		qlen = skb_queue_len(&sd->input_pkt_queue);
3983
		if (qlen)
3984 3985
			skb_queue_splice_tail_init(&sd->input_pkt_queue,
						   &sd->process_queue);
3986

3987
		if (qlen < quota - work) {
E
Eric Dumazet 已提交
3988 3989 3990 3991 3992 3993 3994 3995 3996 3997
			/*
			 * Inline a custom version of __napi_complete().
			 * only current cpu owns and manipulates this napi,
			 * and NAPI_STATE_SCHED is the only possible flag set on backlog.
			 * we can use a plain write instead of clear_bit(),
			 * and we dont need an smp_mb() memory barrier.
			 */
			list_del(&napi->poll_list);
			napi->state = 0;

3998
			quota = work + qlen;
3999
		}
E
Eric Dumazet 已提交
4000
		rps_unlock(sd);
4001 4002
	}
	local_irq_enable();
L
Linus Torvalds 已提交
4003

4004 4005
	return work;
}
L
Linus Torvalds 已提交
4006

4007 4008
/**
 * __napi_schedule - schedule for receive
4009
 * @n: entry to schedule
4010 4011 4012
 *
 * The entry's receive function will be scheduled to run
 */
H
Harvey Harrison 已提交
4013
void __napi_schedule(struct napi_struct *n)
4014 4015
{
	unsigned long flags;
L
Linus Torvalds 已提交
4016

4017
	local_irq_save(flags);
E
Eric Dumazet 已提交
4018
	____napi_schedule(&__get_cpu_var(softnet_data), n);
4019
	local_irq_restore(flags);
L
Linus Torvalds 已提交
4020
}
4021 4022
EXPORT_SYMBOL(__napi_schedule);

4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044
void __napi_complete(struct napi_struct *n)
{
	BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
	BUG_ON(n->gro_list);

	list_del(&n->poll_list);
	smp_mb__before_clear_bit();
	clear_bit(NAPI_STATE_SCHED, &n->state);
}
EXPORT_SYMBOL(__napi_complete);

void napi_complete(struct napi_struct *n)
{
	unsigned long flags;

	/*
	 * don't let napi dequeue from the cpu poll list
	 * just in case its running on a different cpu
	 */
	if (unlikely(test_bit(NAPI_STATE_NPSVC, &n->state)))
		return;

4045
	napi_gro_flush(n, false);
4046 4047 4048 4049 4050 4051 4052 4053 4054 4055
	local_irq_save(flags);
	__napi_complete(n);
	local_irq_restore(flags);
}
EXPORT_SYMBOL(napi_complete);

void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
		    int (*poll)(struct napi_struct *, int), int weight)
{
	INIT_LIST_HEAD(&napi->poll_list);
4056
	napi->gro_count = 0;
4057
	napi->gro_list = NULL;
H
Herbert Xu 已提交
4058
	napi->skb = NULL;
4059
	napi->poll = poll;
E
Eric Dumazet 已提交
4060 4061 4062
	if (weight > NAPI_POLL_WEIGHT)
		pr_err_once("netif_napi_add() called with weight %d on device %s\n",
			    weight, dev->name);
4063 4064 4065
	napi->weight = weight;
	list_add(&napi->dev_list, &dev->napi_list);
	napi->dev = dev;
H
Herbert Xu 已提交
4066
#ifdef CONFIG_NETPOLL
4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077
	spin_lock_init(&napi->poll_lock);
	napi->poll_owner = -1;
#endif
	set_bit(NAPI_STATE_SCHED, &napi->state);
}
EXPORT_SYMBOL(netif_napi_add);

void netif_napi_del(struct napi_struct *napi)
{
	struct sk_buff *skb, *next;

4078
	list_del_init(&napi->dev_list);
4079
	napi_free_frags(napi);
4080 4081 4082 4083 4084 4085 4086 4087

	for (skb = napi->gro_list; skb; skb = next) {
		next = skb->next;
		skb->next = NULL;
		kfree_skb(skb);
	}

	napi->gro_list = NULL;
4088
	napi->gro_count = 0;
4089 4090 4091
}
EXPORT_SYMBOL(netif_napi_del);

L
Linus Torvalds 已提交
4092 4093
static void net_rx_action(struct softirq_action *h)
{
4094
	struct softnet_data *sd = &__get_cpu_var(softnet_data);
4095
	unsigned long time_limit = jiffies + 2;
4096
	int budget = netdev_budget;
4097 4098
	void *have;

L
Linus Torvalds 已提交
4099 4100
	local_irq_disable();

4101
	while (!list_empty(&sd->poll_list)) {
4102 4103
		struct napi_struct *n;
		int work, weight;
L
Linus Torvalds 已提交
4104

4105
		/* If softirq window is exhuasted then punt.
4106 4107
		 * Allow this to run for 2 jiffies since which will allow
		 * an average latency of 1.5/HZ.
4108
		 */
4109
		if (unlikely(budget <= 0 || time_after(jiffies, time_limit)))
L
Linus Torvalds 已提交
4110 4111 4112 4113
			goto softnet_break;

		local_irq_enable();

4114 4115 4116 4117 4118
		/* Even though interrupts have been re-enabled, this
		 * access is safe because interrupts can only add new
		 * entries to the tail of this list, and only ->poll()
		 * calls can remove this head entry from the list.
		 */
4119
		n = list_first_entry(&sd->poll_list, struct napi_struct, poll_list);
L
Linus Torvalds 已提交
4120

4121 4122 4123 4124
		have = netpoll_poll_lock(n);

		weight = n->weight;

4125 4126 4127 4128
		/* This NAPI_STATE_SCHED test is for avoiding a race
		 * with netpoll's poll_napi().  Only the entity which
		 * obtains the lock and sees NAPI_STATE_SCHED set will
		 * actually make the ->poll() call.  Therefore we avoid
L
Lucas De Marchi 已提交
4129
		 * accidentally calling ->poll() when NAPI is not scheduled.
4130 4131
		 */
		work = 0;
4132
		if (test_bit(NAPI_STATE_SCHED, &n->state)) {
4133
			work = n->poll(n, weight);
4134 4135
			trace_napi_poll(n);
		}
4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147

		WARN_ON_ONCE(work > weight);

		budget -= work;

		local_irq_disable();

		/* Drivers must not modify the NAPI state if they
		 * consume the entire weight.  In such cases this code
		 * still "owns" the NAPI instance and therefore can
		 * move the instance around on the list at-will.
		 */
4148
		if (unlikely(work == weight)) {
4149 4150 4151 4152
			if (unlikely(napi_disable_pending(n))) {
				local_irq_enable();
				napi_complete(n);
				local_irq_disable();
4153 4154 4155 4156 4157 4158 4159 4160 4161
			} else {
				if (n->gro_list) {
					/* flush too old packets
					 * If HZ < 1000, flush all packets.
					 */
					local_irq_enable();
					napi_gro_flush(n, HZ >= 1000);
					local_irq_disable();
				}
4162
				list_move_tail(&n->poll_list, &sd->poll_list);
4163
			}
4164
		}
4165 4166

		netpoll_poll_unlock(have);
L
Linus Torvalds 已提交
4167 4168
	}
out:
4169
	net_rps_action_and_irq_enable(sd);
T
Tom Herbert 已提交
4170

4171 4172 4173 4174 4175
#ifdef CONFIG_NET_DMA
	/*
	 * There may not be any more sk_buffs coming right now, so push
	 * any pending DMA copies to hardware
	 */
4176
	dma_issue_pending_all();
4177
#endif
4178

L
Linus Torvalds 已提交
4179 4180 4181
	return;

softnet_break:
C
Changli Gao 已提交
4182
	sd->time_squeeze++;
L
Linus Torvalds 已提交
4183 4184 4185 4186
	__raise_softirq_irqoff(NET_RX_SOFTIRQ);
	goto out;
}

J
Jiri Pirko 已提交
4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412
struct netdev_upper {
	struct net_device *dev;
	bool master;
	struct list_head list;
	struct rcu_head rcu;
	struct list_head search_list;
};

static void __append_search_uppers(struct list_head *search_list,
				   struct net_device *dev)
{
	struct netdev_upper *upper;

	list_for_each_entry(upper, &dev->upper_dev_list, list) {
		/* check if this upper is not already in search list */
		if (list_empty(&upper->search_list))
			list_add_tail(&upper->search_list, search_list);
	}
}

static bool __netdev_search_upper_dev(struct net_device *dev,
				      struct net_device *upper_dev)
{
	LIST_HEAD(search_list);
	struct netdev_upper *upper;
	struct netdev_upper *tmp;
	bool ret = false;

	__append_search_uppers(&search_list, dev);
	list_for_each_entry(upper, &search_list, search_list) {
		if (upper->dev == upper_dev) {
			ret = true;
			break;
		}
		__append_search_uppers(&search_list, upper->dev);
	}
	list_for_each_entry_safe(upper, tmp, &search_list, search_list)
		INIT_LIST_HEAD(&upper->search_list);
	return ret;
}

static struct netdev_upper *__netdev_find_upper(struct net_device *dev,
						struct net_device *upper_dev)
{
	struct netdev_upper *upper;

	list_for_each_entry(upper, &dev->upper_dev_list, list) {
		if (upper->dev == upper_dev)
			return upper;
	}
	return NULL;
}

/**
 * netdev_has_upper_dev - Check if device is linked to an upper device
 * @dev: device
 * @upper_dev: upper device to check
 *
 * Find out if a device is linked to specified upper device and return true
 * in case it is. Note that this checks only immediate upper device,
 * not through a complete stack of devices. The caller must hold the RTNL lock.
 */
bool netdev_has_upper_dev(struct net_device *dev,
			  struct net_device *upper_dev)
{
	ASSERT_RTNL();

	return __netdev_find_upper(dev, upper_dev);
}
EXPORT_SYMBOL(netdev_has_upper_dev);

/**
 * netdev_has_any_upper_dev - Check if device is linked to some device
 * @dev: device
 *
 * Find out if a device is linked to an upper device and return true in case
 * it is. The caller must hold the RTNL lock.
 */
bool netdev_has_any_upper_dev(struct net_device *dev)
{
	ASSERT_RTNL();

	return !list_empty(&dev->upper_dev_list);
}
EXPORT_SYMBOL(netdev_has_any_upper_dev);

/**
 * netdev_master_upper_dev_get - Get master upper device
 * @dev: device
 *
 * Find a master upper device and return pointer to it or NULL in case
 * it's not there. The caller must hold the RTNL lock.
 */
struct net_device *netdev_master_upper_dev_get(struct net_device *dev)
{
	struct netdev_upper *upper;

	ASSERT_RTNL();

	if (list_empty(&dev->upper_dev_list))
		return NULL;

	upper = list_first_entry(&dev->upper_dev_list,
				 struct netdev_upper, list);
	if (likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get);

/**
 * netdev_master_upper_dev_get_rcu - Get master upper device
 * @dev: device
 *
 * Find a master upper device and return pointer to it or NULL in case
 * it's not there. The caller must hold the RCU read lock.
 */
struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev)
{
	struct netdev_upper *upper;

	upper = list_first_or_null_rcu(&dev->upper_dev_list,
				       struct netdev_upper, list);
	if (upper && likely(upper->master))
		return upper->dev;
	return NULL;
}
EXPORT_SYMBOL(netdev_master_upper_dev_get_rcu);

static int __netdev_upper_dev_link(struct net_device *dev,
				   struct net_device *upper_dev, bool master)
{
	struct netdev_upper *upper;

	ASSERT_RTNL();

	if (dev == upper_dev)
		return -EBUSY;

	/* To prevent loops, check if dev is not upper device to upper_dev. */
	if (__netdev_search_upper_dev(upper_dev, dev))
		return -EBUSY;

	if (__netdev_find_upper(dev, upper_dev))
		return -EEXIST;

	if (master && netdev_master_upper_dev_get(dev))
		return -EBUSY;

	upper = kmalloc(sizeof(*upper), GFP_KERNEL);
	if (!upper)
		return -ENOMEM;

	upper->dev = upper_dev;
	upper->master = master;
	INIT_LIST_HEAD(&upper->search_list);

	/* Ensure that master upper link is always the first item in list. */
	if (master)
		list_add_rcu(&upper->list, &dev->upper_dev_list);
	else
		list_add_tail_rcu(&upper->list, &dev->upper_dev_list);
	dev_hold(upper_dev);

	return 0;
}

/**
 * netdev_upper_dev_link - Add a link to the upper device
 * @dev: device
 * @upper_dev: new upper device
 *
 * Adds a link to device which is upper to this one. The caller must hold
 * the RTNL lock. On a failure a negative errno code is returned.
 * On success the reference counts are adjusted and the function
 * returns zero.
 */
int netdev_upper_dev_link(struct net_device *dev,
			  struct net_device *upper_dev)
{
	return __netdev_upper_dev_link(dev, upper_dev, false);
}
EXPORT_SYMBOL(netdev_upper_dev_link);

/**
 * netdev_master_upper_dev_link - Add a master link to the upper device
 * @dev: device
 * @upper_dev: new upper device
 *
 * Adds a link to device which is upper to this one. In this case, only
 * one master upper device can be linked, although other non-master devices
 * might be linked as well. The caller must hold the RTNL lock.
 * On a failure a negative errno code is returned. On success the reference
 * counts are adjusted and the function returns zero.
 */
int netdev_master_upper_dev_link(struct net_device *dev,
				 struct net_device *upper_dev)
{
	return __netdev_upper_dev_link(dev, upper_dev, true);
}
EXPORT_SYMBOL(netdev_master_upper_dev_link);

/**
 * netdev_upper_dev_unlink - Removes a link to upper device
 * @dev: device
 * @upper_dev: new upper device
 *
 * Removes a link to device which is upper to this one. The caller must hold
 * the RTNL lock.
 */
void netdev_upper_dev_unlink(struct net_device *dev,
			     struct net_device *upper_dev)
{
	struct netdev_upper *upper;

	ASSERT_RTNL();

	upper = __netdev_find_upper(dev, upper_dev);
	if (!upper)
		return;
	list_del_rcu(&upper->list);
	dev_put(upper_dev);
	kfree_rcu(upper, rcu);
}
EXPORT_SYMBOL(netdev_upper_dev_unlink);

4413 4414
static void dev_change_rx_flags(struct net_device *dev, int flags)
{
4415 4416 4417 4418
	const struct net_device_ops *ops = dev->netdev_ops;

	if ((dev->flags & IFF_UP) && ops->ndo_change_rx_flags)
		ops->ndo_change_rx_flags(dev, flags);
4419 4420
}

4421
static int __dev_set_promiscuity(struct net_device *dev, int inc)
L
Linus Torvalds 已提交
4422
{
4423
	unsigned int old_flags = dev->flags;
4424 4425
	kuid_t uid;
	kgid_t gid;
L
Linus Torvalds 已提交
4426

4427 4428
	ASSERT_RTNL();

4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439
	dev->flags |= IFF_PROMISC;
	dev->promiscuity += inc;
	if (dev->promiscuity == 0) {
		/*
		 * Avoid overflow.
		 * If inc causes overflow, untouch promisc and return error.
		 */
		if (inc < 0)
			dev->flags &= ~IFF_PROMISC;
		else {
			dev->promiscuity -= inc;
4440 4441
			pr_warn("%s: promiscuity touches roof, set promiscuity failed. promiscuity feature of device might be broken.\n",
				dev->name);
4442 4443 4444
			return -EOVERFLOW;
		}
	}
4445
	if (dev->flags != old_flags) {
4446 4447 4448
		pr_info("device %s %s promiscuous mode\n",
			dev->name,
			dev->flags & IFF_PROMISC ? "entered" : "left");
4449 4450
		if (audit_enabled) {
			current_uid_gid(&uid, &gid);
4451 4452 4453 4454 4455
			audit_log(current->audit_context, GFP_ATOMIC,
				AUDIT_ANOM_PROMISCUOUS,
				"dev=%s prom=%d old_prom=%d auid=%u uid=%u gid=%u ses=%u",
				dev->name, (dev->flags & IFF_PROMISC),
				(old_flags & IFF_PROMISC),
4456
				from_kuid(&init_user_ns, audit_get_loginuid(current)),
4457 4458
				from_kuid(&init_user_ns, uid),
				from_kgid(&init_user_ns, gid),
4459
				audit_get_sessionid(current));
4460
		}
4461

4462
		dev_change_rx_flags(dev, IFF_PROMISC);
L
Linus Torvalds 已提交
4463
	}
4464
	return 0;
L
Linus Torvalds 已提交
4465 4466
}

4467 4468 4469 4470 4471 4472 4473 4474 4475
/**
 *	dev_set_promiscuity	- update promiscuity count on a device
 *	@dev: device
 *	@inc: modifier
 *
 *	Add or remove promiscuity from a device. While the count in the device
 *	remains above zero the interface remains promiscuous. Once it hits zero
 *	the device reverts back to normal filtering operation. A negative inc
 *	value is used to drop promiscuity on the device.
4476
 *	Return 0 if successful or a negative errno code on error.
4477
 */
4478
int dev_set_promiscuity(struct net_device *dev, int inc)
4479
{
4480
	unsigned int old_flags = dev->flags;
4481
	int err;
4482

4483
	err = __dev_set_promiscuity(dev, inc);
4484
	if (err < 0)
4485
		return err;
4486 4487
	if (dev->flags != old_flags)
		dev_set_rx_mode(dev);
4488
	return err;
4489
}
E
Eric Dumazet 已提交
4490
EXPORT_SYMBOL(dev_set_promiscuity);
4491

L
Linus Torvalds 已提交
4492 4493 4494 4495 4496 4497 4498 4499 4500 4501
/**
 *	dev_set_allmulti	- update allmulti count on a device
 *	@dev: device
 *	@inc: modifier
 *
 *	Add or remove reception of all multicast frames to a device. While the
 *	count in the device remains above zero the interface remains listening
 *	to all interfaces. Once it hits zero the device reverts back to normal
 *	filtering operation. A negative @inc value is used to drop the counter
 *	when releasing a resource needing all multicasts.
4502
 *	Return 0 if successful or a negative errno code on error.
L
Linus Torvalds 已提交
4503 4504
 */

4505
int dev_set_allmulti(struct net_device *dev, int inc)
L
Linus Torvalds 已提交
4506
{
4507
	unsigned int old_flags = dev->flags;
L
Linus Torvalds 已提交
4508

4509 4510
	ASSERT_RTNL();

L
Linus Torvalds 已提交
4511
	dev->flags |= IFF_ALLMULTI;
4512 4513 4514 4515 4516 4517 4518 4519 4520 4521
	dev->allmulti += inc;
	if (dev->allmulti == 0) {
		/*
		 * Avoid overflow.
		 * If inc causes overflow, untouch allmulti and return error.
		 */
		if (inc < 0)
			dev->flags &= ~IFF_ALLMULTI;
		else {
			dev->allmulti -= inc;
4522 4523
			pr_warn("%s: allmulti touches roof, set allmulti failed. allmulti feature of device might be broken.\n",
				dev->name);
4524 4525 4526
			return -EOVERFLOW;
		}
	}
4527
	if (dev->flags ^ old_flags) {
4528
		dev_change_rx_flags(dev, IFF_ALLMULTI);
4529
		dev_set_rx_mode(dev);
4530
	}
4531
	return 0;
4532
}
E
Eric Dumazet 已提交
4533
EXPORT_SYMBOL(dev_set_allmulti);
4534 4535 4536 4537

/*
 *	Upload unicast and multicast address lists to device and
 *	configure RX filtering. When the device doesn't support unicast
J
Joe Perches 已提交
4538
 *	filtering it is put in promiscuous mode while unicast addresses
4539 4540 4541 4542
 *	are present.
 */
void __dev_set_rx_mode(struct net_device *dev)
{
4543 4544
	const struct net_device_ops *ops = dev->netdev_ops;

4545 4546 4547 4548 4549
	/* dev_open will call this function so the list will stay sane. */
	if (!(dev->flags&IFF_UP))
		return;

	if (!netif_device_present(dev))
4550
		return;
4551

4552
	if (!(dev->priv_flags & IFF_UNICAST_FLT)) {
4553 4554 4555
		/* Unicast addresses changes may only happen under the rtnl,
		 * therefore calling __dev_set_promiscuity here is safe.
		 */
4556
		if (!netdev_uc_empty(dev) && !dev->uc_promisc) {
4557
			__dev_set_promiscuity(dev, 1);
4558
			dev->uc_promisc = true;
4559
		} else if (netdev_uc_empty(dev) && dev->uc_promisc) {
4560
			__dev_set_promiscuity(dev, -1);
4561
			dev->uc_promisc = false;
4562 4563
		}
	}
4564 4565 4566

	if (ops->ndo_set_rx_mode)
		ops->ndo_set_rx_mode(dev);
4567 4568 4569 4570
}

void dev_set_rx_mode(struct net_device *dev)
{
4571
	netif_addr_lock_bh(dev);
4572
	__dev_set_rx_mode(dev);
4573
	netif_addr_unlock_bh(dev);
L
Linus Torvalds 已提交
4574 4575
}

4576 4577 4578 4579 4580 4581
/**
 *	dev_get_flags - get flags reported to userspace
 *	@dev: device
 *
 *	Get the combination of flag bits exported through APIs to userspace.
 */
4582
unsigned int dev_get_flags(const struct net_device *dev)
L
Linus Torvalds 已提交
4583
{
4584
	unsigned int flags;
L
Linus Torvalds 已提交
4585 4586 4587

	flags = (dev->flags & ~(IFF_PROMISC |
				IFF_ALLMULTI |
S
Stefan Rompf 已提交
4588 4589 4590
				IFF_RUNNING |
				IFF_LOWER_UP |
				IFF_DORMANT)) |
L
Linus Torvalds 已提交
4591 4592 4593
		(dev->gflags & (IFF_PROMISC |
				IFF_ALLMULTI));

S
Stefan Rompf 已提交
4594 4595 4596 4597 4598 4599 4600 4601
	if (netif_running(dev)) {
		if (netif_oper_up(dev))
			flags |= IFF_RUNNING;
		if (netif_carrier_ok(dev))
			flags |= IFF_LOWER_UP;
		if (netif_dormant(dev))
			flags |= IFF_DORMANT;
	}
L
Linus Torvalds 已提交
4602 4603 4604

	return flags;
}
E
Eric Dumazet 已提交
4605
EXPORT_SYMBOL(dev_get_flags);
L
Linus Torvalds 已提交
4606

4607
int __dev_change_flags(struct net_device *dev, unsigned int flags)
L
Linus Torvalds 已提交
4608
{
4609
	unsigned int old_flags = dev->flags;
4610
	int ret;
L
Linus Torvalds 已提交
4611

4612 4613
	ASSERT_RTNL();

L
Linus Torvalds 已提交
4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627
	/*
	 *	Set the flags on our device.
	 */

	dev->flags = (flags & (IFF_DEBUG | IFF_NOTRAILERS | IFF_NOARP |
			       IFF_DYNAMIC | IFF_MULTICAST | IFF_PORTSEL |
			       IFF_AUTOMEDIA)) |
		     (dev->flags & (IFF_UP | IFF_VOLATILE | IFF_PROMISC |
				    IFF_ALLMULTI));

	/*
	 *	Load in the correct multicast list now the flags have changed.
	 */

4628 4629
	if ((old_flags ^ flags) & IFF_MULTICAST)
		dev_change_rx_flags(dev, IFF_MULTICAST);
4630

4631
	dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
4632 4633 4634 4635 4636 4637 4638 4639 4640

	/*
	 *	Have we downed the interface. We handle IFF_UP ourselves
	 *	according to user attempts to set it, rather than blindly
	 *	setting it.
	 */

	ret = 0;
	if ((old_flags ^ flags) & IFF_UP) {	/* Bit is different  ? */
4641
		ret = ((old_flags & IFF_UP) ? __dev_close : __dev_open)(dev);
L
Linus Torvalds 已提交
4642 4643

		if (!ret)
4644
			dev_set_rx_mode(dev);
L
Linus Torvalds 已提交
4645 4646 4647
	}

	if ((flags ^ dev->gflags) & IFF_PROMISC) {
E
Eric Dumazet 已提交
4648 4649
		int inc = (flags & IFF_PROMISC) ? 1 : -1;

L
Linus Torvalds 已提交
4650 4651 4652 4653 4654 4655 4656 4657 4658
		dev->gflags ^= IFF_PROMISC;
		dev_set_promiscuity(dev, inc);
	}

	/* NOTE: order of synchronization of IFF_PROMISC and IFF_ALLMULTI
	   is important. Some (broken) drivers set IFF_PROMISC, when
	   IFF_ALLMULTI is requested not asking us and not reporting.
	 */
	if ((flags ^ dev->gflags) & IFF_ALLMULTI) {
E
Eric Dumazet 已提交
4659 4660
		int inc = (flags & IFF_ALLMULTI) ? 1 : -1;

L
Linus Torvalds 已提交
4661 4662 4663 4664
		dev->gflags ^= IFF_ALLMULTI;
		dev_set_allmulti(dev, inc);
	}

4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691
	return ret;
}

void __dev_notify_flags(struct net_device *dev, unsigned int old_flags)
{
	unsigned int changes = dev->flags ^ old_flags;

	if (changes & IFF_UP) {
		if (dev->flags & IFF_UP)
			call_netdevice_notifiers(NETDEV_UP, dev);
		else
			call_netdevice_notifiers(NETDEV_DOWN, dev);
	}

	if (dev->flags & IFF_UP &&
	    (changes & ~(IFF_UP | IFF_PROMISC | IFF_ALLMULTI | IFF_VOLATILE)))
		call_netdevice_notifiers(NETDEV_CHANGE, dev);
}

/**
 *	dev_change_flags - change device settings
 *	@dev: device
 *	@flags: device state flags
 *
 *	Change settings on device based state flags. The flags are
 *	in the userspace exported format.
 */
4692
int dev_change_flags(struct net_device *dev, unsigned int flags)
4693
{
4694 4695
	int ret;
	unsigned int changes, old_flags = dev->flags;
4696 4697 4698 4699 4700 4701

	ret = __dev_change_flags(dev, flags);
	if (ret < 0)
		return ret;

	changes = old_flags ^ dev->flags;
4702 4703
	if (changes)
		rtmsg_ifinfo(RTM_NEWLINK, dev, changes);
L
Linus Torvalds 已提交
4704

4705
	__dev_notify_flags(dev, old_flags);
L
Linus Torvalds 已提交
4706 4707
	return ret;
}
E
Eric Dumazet 已提交
4708
EXPORT_SYMBOL(dev_change_flags);
L
Linus Torvalds 已提交
4709

4710 4711 4712 4713 4714 4715 4716
/**
 *	dev_set_mtu - Change maximum transfer unit
 *	@dev: device
 *	@new_mtu: new transfer unit
 *
 *	Change the maximum transfer size of the network device.
 */
L
Linus Torvalds 已提交
4717 4718
int dev_set_mtu(struct net_device *dev, int new_mtu)
{
4719
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732
	int err;

	if (new_mtu == dev->mtu)
		return 0;

	/*	MTU must be positive.	 */
	if (new_mtu < 0)
		return -EINVAL;

	if (!netif_device_present(dev))
		return -ENODEV;

	err = 0;
4733 4734
	if (ops->ndo_change_mtu)
		err = ops->ndo_change_mtu(dev, new_mtu);
L
Linus Torvalds 已提交
4735 4736
	else
		dev->mtu = new_mtu;
4737

4738
	if (!err)
4739
		call_netdevice_notifiers(NETDEV_CHANGEMTU, dev);
L
Linus Torvalds 已提交
4740 4741
	return err;
}
E
Eric Dumazet 已提交
4742
EXPORT_SYMBOL(dev_set_mtu);
L
Linus Torvalds 已提交
4743

4744 4745 4746 4747 4748 4749 4750 4751 4752 4753 4754
/**
 *	dev_set_group - Change group this device belongs to
 *	@dev: device
 *	@new_group: group this device should belong to
 */
void dev_set_group(struct net_device *dev, int new_group)
{
	dev->group = new_group;
}
EXPORT_SYMBOL(dev_set_group);

4755 4756 4757 4758 4759 4760 4761
/**
 *	dev_set_mac_address - Change Media Access Control Address
 *	@dev: device
 *	@sa: new address
 *
 *	Change the hardware (MAC) address of the device
 */
L
Linus Torvalds 已提交
4762 4763
int dev_set_mac_address(struct net_device *dev, struct sockaddr *sa)
{
4764
	const struct net_device_ops *ops = dev->netdev_ops;
L
Linus Torvalds 已提交
4765 4766
	int err;

4767
	if (!ops->ndo_set_mac_address)
L
Linus Torvalds 已提交
4768 4769 4770 4771 4772
		return -EOPNOTSUPP;
	if (sa->sa_family != dev->type)
		return -EINVAL;
	if (!netif_device_present(dev))
		return -ENODEV;
4773
	err = ops->ndo_set_mac_address(dev, sa);
4774 4775
	if (err)
		return err;
J
Jiri Pirko 已提交
4776
	dev->addr_assign_type = NET_ADDR_SET;
4777
	call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
4778
	add_device_randomness(dev->dev_addr, dev->addr_len);
4779
	return 0;
L
Linus Torvalds 已提交
4780
}
E
Eric Dumazet 已提交
4781
EXPORT_SYMBOL(dev_set_mac_address);
L
Linus Torvalds 已提交
4782

J
Jiri Pirko 已提交
4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801
/**
 *	dev_change_carrier - Change device carrier
 *	@dev: device
 *	@new_carries: new value
 *
 *	Change device carrier
 */
int dev_change_carrier(struct net_device *dev, bool new_carrier)
{
	const struct net_device_ops *ops = dev->netdev_ops;

	if (!ops->ndo_change_carrier)
		return -EOPNOTSUPP;
	if (!netif_device_present(dev))
		return -ENODEV;
	return ops->ndo_change_carrier(dev, new_carrier);
}
EXPORT_SYMBOL(dev_change_carrier);

L
Linus Torvalds 已提交
4802 4803
/**
 *	dev_new_index	-	allocate an ifindex
4804
 *	@net: the applicable net namespace
L
Linus Torvalds 已提交
4805 4806 4807 4808 4809
 *
 *	Returns a suitable unique value for a new device interface
 *	number.  The caller must hold the rtnl semaphore or the
 *	dev_base_lock to be sure it remains unique.
 */
4810
static int dev_new_index(struct net *net)
L
Linus Torvalds 已提交
4811
{
4812
	int ifindex = net->ifindex;
L
Linus Torvalds 已提交
4813 4814 4815
	for (;;) {
		if (++ifindex <= 0)
			ifindex = 1;
4816
		if (!__dev_get_by_index(net, ifindex))
4817
			return net->ifindex = ifindex;
L
Linus Torvalds 已提交
4818 4819 4820 4821
	}
}

/* Delayed registration/unregisteration */
4822
static LIST_HEAD(net_todo_list);
L
Linus Torvalds 已提交
4823

4824
static void net_set_todo(struct net_device *dev)
L
Linus Torvalds 已提交
4825 4826 4827 4828
{
	list_add_tail(&dev->todo_list, &net_todo_list);
}

4829
static void rollback_registered_many(struct list_head *head)
4830
{
4831
	struct net_device *dev, *tmp;
4832

4833 4834 4835
	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

4836
	list_for_each_entry_safe(dev, tmp, head, unreg_list) {
4837
		/* Some devices call without registering
4838 4839
		 * for initialization unwind. Remove those
		 * devices and proceed with the remaining.
4840 4841
		 */
		if (dev->reg_state == NETREG_UNINITIALIZED) {
4842 4843
			pr_debug("unregister_netdevice: device %s/%p never was registered\n",
				 dev->name, dev);
4844

4845
			WARN_ON(1);
4846 4847
			list_del(&dev->unreg_list);
			continue;
4848
		}
4849
		dev->dismantle = true;
4850
		BUG_ON(dev->reg_state != NETREG_REGISTERED);
4851
	}
4852

4853 4854
	/* If device is running, close it first. */
	dev_close_many(head);
4855

4856
	list_for_each_entry(dev, head, unreg_list) {
4857 4858
		/* And unlink it from device chain. */
		unlist_netdevice(dev);
4859

4860 4861
		dev->reg_state = NETREG_UNREGISTERING;
	}
4862 4863 4864

	synchronize_net();

4865 4866 4867
	list_for_each_entry(dev, head, unreg_list) {
		/* Shutdown queueing discipline. */
		dev_shutdown(dev);
4868 4869


4870 4871 4872 4873
		/* Notify protocols, that we are about to destroy
		   this device. They should clean all the things.
		*/
		call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
4874

4875 4876 4877 4878
		if (!dev->rtnl_link_ops ||
		    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
			rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);

4879 4880 4881
		/*
		 *	Flush the unicast and multicast chains
		 */
4882
		dev_uc_flush(dev);
4883
		dev_mc_flush(dev);
4884

4885 4886
		if (dev->netdev_ops->ndo_uninit)
			dev->netdev_ops->ndo_uninit(dev);
4887

J
Jiri Pirko 已提交
4888 4889
		/* Notifier chain MUST detach us all upper devices. */
		WARN_ON(netdev_has_any_upper_dev(dev));
4890

4891 4892
		/* Remove entries from kobject tree */
		netdev_unregister_kobject(dev);
4893 4894 4895 4896
#ifdef CONFIG_XPS
		/* Remove XPS queueing entries */
		netif_reset_xps_queues_gt(dev, 0);
#endif
4897
	}
4898

4899
	synchronize_net();
4900

4901
	list_for_each_entry(dev, head, unreg_list)
4902 4903 4904 4905 4906 4907 4908 4909 4910
		dev_put(dev);
}

static void rollback_registered(struct net_device *dev)
{
	LIST_HEAD(single);

	list_add(&dev->unreg_list, &single);
	rollback_registered_many(&single);
E
Eric Dumazet 已提交
4911
	list_del(&single);
4912 4913
}

4914 4915
static netdev_features_t netdev_fix_features(struct net_device *dev,
	netdev_features_t features)
4916
{
4917 4918 4919
	/* Fix illegal checksum combinations */
	if ((features & NETIF_F_HW_CSUM) &&
	    (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
4920
		netdev_warn(dev, "mixed HW and IP checksum settings.\n");
4921 4922 4923
		features &= ~(NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
	}

4924 4925 4926
	/* Fix illegal SG+CSUM combinations. */
	if ((features & NETIF_F_SG) &&
	    !(features & NETIF_F_ALL_CSUM)) {
4927 4928
		netdev_dbg(dev,
			"Dropping NETIF_F_SG since no checksum feature.\n");
4929 4930 4931 4932
		features &= ~NETIF_F_SG;
	}

	/* TSO requires that SG is present as well. */
4933
	if ((features & NETIF_F_ALL_TSO) && !(features & NETIF_F_SG)) {
4934
		netdev_dbg(dev, "Dropping TSO features since no SG feature.\n");
4935
		features &= ~NETIF_F_ALL_TSO;
4936 4937
	}

4938 4939 4940 4941
	/* TSO ECN requires that TSO is present as well. */
	if ((features & NETIF_F_ALL_TSO) == NETIF_F_TSO_ECN)
		features &= ~NETIF_F_TSO_ECN;

4942 4943
	/* Software GSO depends on SG. */
	if ((features & NETIF_F_GSO) && !(features & NETIF_F_SG)) {
4944
		netdev_dbg(dev, "Dropping NETIF_F_GSO since no SG feature.\n");
4945 4946 4947
		features &= ~NETIF_F_GSO;
	}

4948
	/* UFO needs SG and checksumming */
4949
	if (features & NETIF_F_UFO) {
4950 4951 4952 4953
		/* maybe split UFO into V4 and V6? */
		if (!((features & NETIF_F_GEN_CSUM) ||
		    (features & (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))
			    == (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM))) {
4954
			netdev_dbg(dev,
4955
				"Dropping NETIF_F_UFO since no checksum offload features.\n");
4956 4957 4958 4959
			features &= ~NETIF_F_UFO;
		}

		if (!(features & NETIF_F_SG)) {
4960
			netdev_dbg(dev,
4961
				"Dropping NETIF_F_UFO since no NETIF_F_SG feature.\n");
4962 4963 4964 4965 4966 4967 4968
			features &= ~NETIF_F_UFO;
		}
	}

	return features;
}

4969
int __netdev_update_features(struct net_device *dev)
4970
{
4971
	netdev_features_t features;
4972 4973
	int err = 0;

4974 4975
	ASSERT_RTNL();

4976 4977 4978 4979 4980 4981 4982 4983 4984
	features = netdev_get_wanted_features(dev);

	if (dev->netdev_ops->ndo_fix_features)
		features = dev->netdev_ops->ndo_fix_features(dev, features);

	/* driver might be less strict about feature dependencies */
	features = netdev_fix_features(dev, features);

	if (dev->features == features)
4985
		return 0;
4986

4987 4988
	netdev_dbg(dev, "Features changed: %pNF -> %pNF\n",
		&dev->features, &features);
4989 4990 4991 4992

	if (dev->netdev_ops->ndo_set_features)
		err = dev->netdev_ops->ndo_set_features(dev, features);

4993
	if (unlikely(err < 0)) {
4994
		netdev_err(dev,
4995 4996
			"set_features() failed (%d); wanted %pNF, left %pNF\n",
			err, &features, &dev->features);
4997 4998 4999 5000 5001 5002 5003 5004 5005
		return -1;
	}

	if (!err)
		dev->features = features;

	return 1;
}

5006 5007 5008 5009 5010 5011 5012 5013
/**
 *	netdev_update_features - recalculate device features
 *	@dev: the device to check
 *
 *	Recalculate dev->features set and send notifications if it
 *	has changed. Should be called after driver or hardware dependent
 *	conditions might have changed that influence the features.
 */
5014 5015 5016 5017
void netdev_update_features(struct net_device *dev)
{
	if (__netdev_update_features(dev))
		netdev_features_change(dev);
5018 5019 5020
}
EXPORT_SYMBOL(netdev_update_features);

5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037
/**
 *	netdev_change_features - recalculate device features
 *	@dev: the device to check
 *
 *	Recalculate dev->features set and send notifications even
 *	if they have not changed. Should be called instead of
 *	netdev_update_features() if also dev->vlan_features might
 *	have changed to allow the changes to be propagated to stacked
 *	VLAN devices.
 */
void netdev_change_features(struct net_device *dev)
{
	__netdev_update_features(dev);
	netdev_features_change(dev);
}
EXPORT_SYMBOL(netdev_change_features);

5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064
/**
 *	netif_stacked_transfer_operstate -	transfer operstate
 *	@rootdev: the root or lower level device to transfer state from
 *	@dev: the device to transfer operstate to
 *
 *	Transfer operational state from root to device. This is normally
 *	called when a stacking relationship exists between the root
 *	device and the device(a leaf device).
 */
void netif_stacked_transfer_operstate(const struct net_device *rootdev,
					struct net_device *dev)
{
	if (rootdev->operstate == IF_OPER_DORMANT)
		netif_dormant_on(dev);
	else
		netif_dormant_off(dev);

	if (netif_carrier_ok(rootdev)) {
		if (!netif_carrier_ok(dev))
			netif_carrier_on(dev);
	} else {
		if (netif_carrier_ok(dev))
			netif_carrier_off(dev);
	}
}
EXPORT_SYMBOL(netif_stacked_transfer_operstate);

T
Tom Herbert 已提交
5065
#ifdef CONFIG_RPS
5066 5067 5068
static int netif_alloc_rx_queues(struct net_device *dev)
{
	unsigned int i, count = dev->num_rx_queues;
T
Tom Herbert 已提交
5069
	struct netdev_rx_queue *rx;
5070

T
Tom Herbert 已提交
5071
	BUG_ON(count < 1);
5072

T
Tom Herbert 已提交
5073
	rx = kcalloc(count, sizeof(struct netdev_rx_queue), GFP_KERNEL);
5074
	if (!rx)
T
Tom Herbert 已提交
5075
		return -ENOMEM;
5076

T
Tom Herbert 已提交
5077 5078 5079
	dev->_rx = rx;

	for (i = 0; i < count; i++)
T
Tom Herbert 已提交
5080
		rx[i].dev = dev;
5081 5082
	return 0;
}
T
Tom Herbert 已提交
5083
#endif
5084

C
Changli Gao 已提交
5085 5086 5087 5088 5089 5090 5091
static void netdev_init_one_queue(struct net_device *dev,
				  struct netdev_queue *queue, void *_unused)
{
	/* Initialize queue lock */
	spin_lock_init(&queue->_xmit_lock);
	netdev_set_xmit_lockdep_class(&queue->_xmit_lock, dev->type);
	queue->xmit_lock_owner = -1;
5092
	netdev_queue_numa_node_write(queue, NUMA_NO_NODE);
C
Changli Gao 已提交
5093
	queue->dev = dev;
T
Tom Herbert 已提交
5094 5095 5096
#ifdef CONFIG_BQL
	dql_init(&queue->dql, HZ);
#endif
C
Changli Gao 已提交
5097 5098
}

5099 5100 5101 5102 5103 5104 5105 5106
static int netif_alloc_netdev_queues(struct net_device *dev)
{
	unsigned int count = dev->num_tx_queues;
	struct netdev_queue *tx;

	BUG_ON(count < 1);

	tx = kcalloc(count, sizeof(struct netdev_queue), GFP_KERNEL);
5107
	if (!tx)
5108
		return -ENOMEM;
5109

5110
	dev->_tx = tx;
T
Tom Herbert 已提交
5111

5112 5113
	netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL);
	spin_lock_init(&dev->tx_global_lock);
C
Changli Gao 已提交
5114 5115

	return 0;
5116 5117
}

L
Linus Torvalds 已提交
5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137
/**
 *	register_netdevice	- register a network device
 *	@dev: device to register
 *
 *	Take a completed network device structure and add it to the kernel
 *	interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier
 *	chain. 0 is returned on success. A negative errno code is returned
 *	on a failure to set up the device, or if the name is a duplicate.
 *
 *	Callers must hold the rtnl semaphore. You may want
 *	register_netdev() instead of this.
 *
 *	BUGS:
 *	The locking appears insufficient to guarantee two parallel registers
 *	will not get the same name.
 */

int register_netdevice(struct net_device *dev)
{
	int ret;
5138
	struct net *net = dev_net(dev);
L
Linus Torvalds 已提交
5139 5140 5141 5142

	BUG_ON(dev_boot_phase);
	ASSERT_RTNL();

5143 5144
	might_sleep();

L
Linus Torvalds 已提交
5145 5146
	/* When net_device's are persistent, this will be fatal. */
	BUG_ON(dev->reg_state != NETREG_UNINITIALIZED);
5147
	BUG_ON(!net);
L
Linus Torvalds 已提交
5148

5149
	spin_lock_init(&dev->addr_list_lock);
5150
	netdev_set_addr_lockdep_class(dev);
L
Linus Torvalds 已提交
5151 5152 5153

	dev->iflink = -1;

5154
	ret = dev_get_valid_name(net, dev, dev->name);
5155 5156 5157
	if (ret < 0)
		goto out;

L
Linus Torvalds 已提交
5158
	/* Init, if this function is available */
5159 5160
	if (dev->netdev_ops->ndo_init) {
		ret = dev->netdev_ops->ndo_init(dev);
L
Linus Torvalds 已提交
5161 5162 5163
		if (ret) {
			if (ret > 0)
				ret = -EIO;
5164
			goto out;
L
Linus Torvalds 已提交
5165 5166
		}
	}
5167

5168 5169 5170 5171 5172 5173 5174 5175
	if (((dev->hw_features | dev->features) & NETIF_F_HW_VLAN_FILTER) &&
	    (!dev->netdev_ops->ndo_vlan_rx_add_vid ||
	     !dev->netdev_ops->ndo_vlan_rx_kill_vid)) {
		netdev_WARN(dev, "Buggy VLAN acceleration in driver!\n");
		ret = -EINVAL;
		goto err_uninit;
	}

5176 5177 5178 5179 5180 5181
	ret = -EBUSY;
	if (!dev->ifindex)
		dev->ifindex = dev_new_index(net);
	else if (__dev_get_by_index(net, dev->ifindex))
		goto err_uninit;

L
Linus Torvalds 已提交
5182 5183 5184
	if (dev->iflink == -1)
		dev->iflink = dev->ifindex;

5185 5186 5187 5188
	/* Transfer changeable features to wanted_features and enable
	 * software offloads (GSO and GRO).
	 */
	dev->hw_features |= NETIF_F_SOFT_FEATURES;
5189 5190
	dev->features |= NETIF_F_SOFT_FEATURES;
	dev->wanted_features = dev->features & dev->hw_features;
L
Linus Torvalds 已提交
5191

5192
	/* Turn on no cache copy if HW is doing checksum */
5193 5194 5195 5196 5197 5198
	if (!(dev->flags & IFF_LOOPBACK)) {
		dev->hw_features |= NETIF_F_NOCACHE_COPY;
		if (dev->features & NETIF_F_ALL_CSUM) {
			dev->wanted_features |= NETIF_F_NOCACHE_COPY;
			dev->features |= NETIF_F_NOCACHE_COPY;
		}
5199 5200
	}

5201
	/* Make NETIF_F_HIGHDMA inheritable to VLAN devices.
5202
	 */
5203
	dev->vlan_features |= NETIF_F_HIGHDMA;
5204

5205 5206 5207 5208 5209
	ret = call_netdevice_notifiers(NETDEV_POST_INIT, dev);
	ret = notifier_to_errno(ret);
	if (ret)
		goto err_uninit;

5210
	ret = netdev_register_kobject(dev);
5211
	if (ret)
5212
		goto err_uninit;
5213 5214
	dev->reg_state = NETREG_REGISTERED;

5215
	__netdev_update_features(dev);
5216

L
Linus Torvalds 已提交
5217 5218 5219 5220 5221 5222 5223
	/*
	 *	Default initial state at registry is that the
	 *	device is present.
	 */

	set_bit(__LINK_STATE_PRESENT, &dev->state);

5224 5225
	linkwatch_init_dev(dev);

L
Linus Torvalds 已提交
5226 5227
	dev_init_scheduler(dev);
	dev_hold(dev);
5228
	list_netdevice(dev);
5229
	add_device_randomness(dev->dev_addr, dev->addr_len);
L
Linus Torvalds 已提交
5230

5231 5232 5233 5234 5235 5236 5237
	/* If the device has permanent device address, driver should
	 * set dev_addr and also addr_assign_type should be set to
	 * NET_ADDR_PERM (default value).
	 */
	if (dev->addr_assign_type == NET_ADDR_PERM)
		memcpy(dev->perm_addr, dev->dev_addr, dev->addr_len);

L
Linus Torvalds 已提交
5238
	/* Notify protocols, that a new device appeared. */
5239
	ret = call_netdevice_notifiers(NETDEV_REGISTER, dev);
5240
	ret = notifier_to_errno(ret);
5241 5242 5243 5244
	if (ret) {
		rollback_registered(dev);
		dev->reg_state = NETREG_UNREGISTERED;
	}
5245 5246 5247 5248
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
5249 5250 5251
	if (!dev->rtnl_link_ops ||
	    dev->rtnl_link_state == RTNL_LINK_INITIALIZED)
		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
L
Linus Torvalds 已提交
5252 5253 5254

out:
	return ret;
5255 5256

err_uninit:
5257 5258
	if (dev->netdev_ops->ndo_uninit)
		dev->netdev_ops->ndo_uninit(dev);
5259
	goto out;
L
Linus Torvalds 已提交
5260
}
E
Eric Dumazet 已提交
5261
EXPORT_SYMBOL(register_netdevice);
L
Linus Torvalds 已提交
5262

5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293
/**
 *	init_dummy_netdev	- init a dummy network device for NAPI
 *	@dev: device to init
 *
 *	This takes a network device structure and initialize the minimum
 *	amount of fields so it can be used to schedule NAPI polls without
 *	registering a full blown interface. This is to be used by drivers
 *	that need to tie several hardware interfaces to a single NAPI
 *	poll scheduler due to HW limitations.
 */
int init_dummy_netdev(struct net_device *dev)
{
	/* Clear everything. Note we don't initialize spinlocks
	 * are they aren't supposed to be taken by any of the
	 * NAPI code and this dummy netdev is supposed to be
	 * only ever used for NAPI polls
	 */
	memset(dev, 0, sizeof(struct net_device));

	/* make sure we BUG if trying to hit standard
	 * register/unregister code path
	 */
	dev->reg_state = NETREG_DUMMY;

	/* NAPI wants this */
	INIT_LIST_HEAD(&dev->napi_list);

	/* a dummy interface is started by default */
	set_bit(__LINK_STATE_PRESENT, &dev->state);
	set_bit(__LINK_STATE_START, &dev->state);

E
Eric Dumazet 已提交
5294 5295 5296 5297 5298
	/* Note : We dont allocate pcpu_refcnt for dummy devices,
	 * because users of this 'device' dont need to change
	 * its refcount.
	 */

5299 5300 5301 5302 5303
	return 0;
}
EXPORT_SYMBOL_GPL(init_dummy_netdev);


L
Linus Torvalds 已提交
5304 5305 5306 5307 5308 5309 5310 5311 5312
/**
 *	register_netdev	- register a network device
 *	@dev: device to register
 *
 *	Take a completed network device structure and add it to the kernel
 *	interfaces. A %NETDEV_REGISTER message is sent to the netdev notifier
 *	chain. 0 is returned on success. A negative errno code is returned
 *	on a failure to set up the device, or if the name is a duplicate.
 *
5313
 *	This is a wrapper around register_netdevice that takes the rtnl semaphore
L
Linus Torvalds 已提交
5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327
 *	and expands the device name if you passed a format string to
 *	alloc_netdev.
 */
int register_netdev(struct net_device *dev)
{
	int err;

	rtnl_lock();
	err = register_netdevice(dev);
	rtnl_unlock();
	return err;
}
EXPORT_SYMBOL(register_netdev);

E
Eric Dumazet 已提交
5328 5329 5330 5331 5332 5333 5334 5335 5336 5337
int netdev_refcnt_read(const struct net_device *dev)
{
	int i, refcnt = 0;

	for_each_possible_cpu(i)
		refcnt += *per_cpu_ptr(dev->pcpu_refcnt, i);
	return refcnt;
}
EXPORT_SYMBOL(netdev_refcnt_read);

5338
/**
L
Linus Torvalds 已提交
5339
 * netdev_wait_allrefs - wait until all references are gone.
5340
 * @dev: target net_device
L
Linus Torvalds 已提交
5341 5342 5343 5344 5345 5346 5347
 *
 * This is called when unregistering network devices.
 *
 * Any protocol or device that holds a reference should register
 * for netdevice notification, and cleanup and put back the
 * reference if they receive an UNREGISTER event.
 * We can get stuck here if buggy protocols don't correctly
5348
 * call dev_put.
L
Linus Torvalds 已提交
5349 5350 5351 5352
 */
static void netdev_wait_allrefs(struct net_device *dev)
{
	unsigned long rebroadcast_time, warning_time;
E
Eric Dumazet 已提交
5353
	int refcnt;
L
Linus Torvalds 已提交
5354

5355 5356
	linkwatch_forget_dev(dev);

L
Linus Torvalds 已提交
5357
	rebroadcast_time = warning_time = jiffies;
E
Eric Dumazet 已提交
5358 5359 5360
	refcnt = netdev_refcnt_read(dev);

	while (refcnt != 0) {
L
Linus Torvalds 已提交
5361
		if (time_after(jiffies, rebroadcast_time + 1 * HZ)) {
5362
			rtnl_lock();
L
Linus Torvalds 已提交
5363 5364

			/* Rebroadcast unregister notification */
5365
			call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
L
Linus Torvalds 已提交
5366

5367
			__rtnl_unlock();
5368
			rcu_barrier();
5369 5370
			rtnl_lock();

5371
			call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
L
Linus Torvalds 已提交
5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382
			if (test_bit(__LINK_STATE_LINKWATCH_PENDING,
				     &dev->state)) {
				/* We must not have linkwatch events
				 * pending on unregister. If this
				 * happens, we simply run the queue
				 * unscheduled, resulting in a noop
				 * for this device.
				 */
				linkwatch_run_queue();
			}

5383
			__rtnl_unlock();
L
Linus Torvalds 已提交
5384 5385 5386 5387 5388 5389

			rebroadcast_time = jiffies;
		}

		msleep(250);

E
Eric Dumazet 已提交
5390 5391
		refcnt = netdev_refcnt_read(dev);

L
Linus Torvalds 已提交
5392
		if (time_after(jiffies, warning_time + 10 * HZ)) {
5393 5394
			pr_emerg("unregister_netdevice: waiting for %s to become free. Usage count = %d\n",
				 dev->name, refcnt);
L
Linus Torvalds 已提交
5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413
			warning_time = jiffies;
		}
	}
}

/* The sequence is:
 *
 *	rtnl_lock();
 *	...
 *	register_netdevice(x1);
 *	register_netdevice(x2);
 *	...
 *	unregister_netdevice(y1);
 *	unregister_netdevice(y2);
 *      ...
 *	rtnl_unlock();
 *	free_netdev(y1);
 *	free_netdev(y2);
 *
H
Herbert Xu 已提交
5414
 * We are invoked by rtnl_unlock().
L
Linus Torvalds 已提交
5415
 * This allows us to deal with problems:
5416
 * 1) We can delete sysfs objects which invoke hotplug
L
Linus Torvalds 已提交
5417 5418 5419
 *    without deadlocking with linkwatch via keventd.
 * 2) Since we run with the RTNL semaphore not held, we can sleep
 *    safely in order to wait for the netdev refcnt to drop to zero.
H
Herbert Xu 已提交
5420 5421 5422
 *
 * We must not return until all unregister events added during
 * the interval the lock was held have been completed.
L
Linus Torvalds 已提交
5423 5424 5425
 */
void netdev_run_todo(void)
{
5426
	struct list_head list;
L
Linus Torvalds 已提交
5427 5428

	/* Snapshot list, allow later requests */
5429
	list_replace_init(&net_todo_list, &list);
H
Herbert Xu 已提交
5430 5431

	__rtnl_unlock();
5432

5433 5434

	/* Wait for rcu callbacks to finish before next phase */
5435 5436 5437
	if (!list_empty(&list))
		rcu_barrier();

L
Linus Torvalds 已提交
5438 5439
	while (!list_empty(&list)) {
		struct net_device *dev
5440
			= list_first_entry(&list, struct net_device, todo_list);
L
Linus Torvalds 已提交
5441 5442
		list_del(&dev->todo_list);

5443
		rtnl_lock();
5444
		call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
5445
		__rtnl_unlock();
5446

5447
		if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
5448
			pr_err("network todo '%s' but state %d\n",
5449 5450 5451 5452
			       dev->name, dev->reg_state);
			dump_stack();
			continue;
		}
L
Linus Torvalds 已提交
5453

5454
		dev->reg_state = NETREG_UNREGISTERED;
L
Linus Torvalds 已提交
5455

5456
		on_each_cpu(flush_backlog, dev, 1);
5457

5458
		netdev_wait_allrefs(dev);
L
Linus Torvalds 已提交
5459

5460
		/* paranoia */
E
Eric Dumazet 已提交
5461
		BUG_ON(netdev_refcnt_read(dev));
5462 5463
		WARN_ON(rcu_access_pointer(dev->ip_ptr));
		WARN_ON(rcu_access_pointer(dev->ip6_ptr));
5464
		WARN_ON(dev->dn_ptr);
L
Linus Torvalds 已提交
5465

5466 5467
		if (dev->destructor)
			dev->destructor(dev);
5468 5469 5470

		/* Free network device */
		kobject_put(&dev->dev.kobj);
L
Linus Torvalds 已提交
5471 5472 5473
	}
}

5474 5475 5476
/* Convert net_device_stats to rtnl_link_stats64.  They have the same
 * fields in the same order, with only the type differing.
 */
5477 5478
void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
			     const struct net_device_stats *netdev_stats)
5479 5480
{
#if BITS_PER_LONG == 64
5481 5482
	BUILD_BUG_ON(sizeof(*stats64) != sizeof(*netdev_stats));
	memcpy(stats64, netdev_stats, sizeof(*stats64));
5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493
#else
	size_t i, n = sizeof(*stats64) / sizeof(u64);
	const unsigned long *src = (const unsigned long *)netdev_stats;
	u64 *dst = (u64 *)stats64;

	BUILD_BUG_ON(sizeof(*netdev_stats) / sizeof(unsigned long) !=
		     sizeof(*stats64) / sizeof(u64));
	for (i = 0; i < n; i++)
		dst[i] = src[i];
#endif
}
5494
EXPORT_SYMBOL(netdev_stats_to_stats64);
5495

5496 5497 5498
/**
 *	dev_get_stats	- get network device statistics
 *	@dev: device to get statistics from
5499
 *	@storage: place to store stats
5500
 *
5501 5502 5503 5504
 *	Get network statistics from device. Return @storage.
 *	The device driver may provide its own method by setting
 *	dev->netdev_ops->get_stats64 or dev->netdev_ops->get_stats;
 *	otherwise the internal statistics structure is used.
5505
 */
5506 5507
struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
					struct rtnl_link_stats64 *storage)
5508
{
5509 5510
	const struct net_device_ops *ops = dev->netdev_ops;

5511 5512
	if (ops->ndo_get_stats64) {
		memset(storage, 0, sizeof(*storage));
5513 5514
		ops->ndo_get_stats64(dev, storage);
	} else if (ops->ndo_get_stats) {
5515
		netdev_stats_to_stats64(storage, ops->ndo_get_stats(dev));
5516 5517
	} else {
		netdev_stats_to_stats64(storage, &dev->stats);
5518
	}
5519
	storage->rx_dropped += atomic_long_read(&dev->rx_dropped);
5520
	return storage;
R
Rusty Russell 已提交
5521
}
5522
EXPORT_SYMBOL(dev_get_stats);
R
Rusty Russell 已提交
5523

5524
struct netdev_queue *dev_ingress_queue_create(struct net_device *dev)
5525
{
5526
	struct netdev_queue *queue = dev_ingress_queue(dev);
5527

5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539
#ifdef CONFIG_NET_CLS_ACT
	if (queue)
		return queue;
	queue = kzalloc(sizeof(*queue), GFP_KERNEL);
	if (!queue)
		return NULL;
	netdev_init_one_queue(dev, queue, NULL);
	queue->qdisc = &noop_qdisc;
	queue->qdisc_sleeping = &noop_qdisc;
	rcu_assign_pointer(dev->ingress_queue, queue);
#endif
	return queue;
5540 5541
}

5542 5543
static const struct ethtool_ops default_ethtool_ops;

5544 5545 5546 5547 5548 5549 5550 5551
void netdev_set_default_ethtool_ops(struct net_device *dev,
				    const struct ethtool_ops *ops)
{
	if (dev->ethtool_ops == &default_ethtool_ops)
		dev->ethtool_ops = ops;
}
EXPORT_SYMBOL_GPL(netdev_set_default_ethtool_ops);

L
Linus Torvalds 已提交
5552
/**
T
Tom Herbert 已提交
5553
 *	alloc_netdev_mqs - allocate network device
L
Linus Torvalds 已提交
5554 5555 5556
 *	@sizeof_priv:	size of private data to allocate space for
 *	@name:		device name format string
 *	@setup:		callback to initialize device
T
Tom Herbert 已提交
5557 5558
 *	@txqs:		the number of TX subqueues to allocate
 *	@rxqs:		the number of RX subqueues to allocate
L
Linus Torvalds 已提交
5559 5560
 *
 *	Allocates a struct net_device with private data area for driver use
5561
 *	and performs basic initialization.  Also allocates subquue structs
T
Tom Herbert 已提交
5562
 *	for each queue on the device.
L
Linus Torvalds 已提交
5563
 */
T
Tom Herbert 已提交
5564 5565 5566
struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
		void (*setup)(struct net_device *),
		unsigned int txqs, unsigned int rxqs)
L
Linus Torvalds 已提交
5567 5568
{
	struct net_device *dev;
5569
	size_t alloc_size;
5570
	struct net_device *p;
L
Linus Torvalds 已提交
5571

5572 5573
	BUG_ON(strlen(name) >= sizeof(dev->name));

T
Tom Herbert 已提交
5574
	if (txqs < 1) {
5575
		pr_err("alloc_netdev: Unable to allocate device with zero queues\n");
5576 5577 5578
		return NULL;
	}

T
Tom Herbert 已提交
5579 5580
#ifdef CONFIG_RPS
	if (rxqs < 1) {
5581
		pr_err("alloc_netdev: Unable to allocate device with zero RX queues\n");
T
Tom Herbert 已提交
5582 5583 5584 5585
		return NULL;
	}
#endif

5586
	alloc_size = sizeof(struct net_device);
5587 5588
	if (sizeof_priv) {
		/* ensure 32-byte alignment of private area */
5589
		alloc_size = ALIGN(alloc_size, NETDEV_ALIGN);
5590 5591 5592
		alloc_size += sizeof_priv;
	}
	/* ensure 32-byte alignment of whole construct */
5593
	alloc_size += NETDEV_ALIGN - 1;
L
Linus Torvalds 已提交
5594

5595
	p = kzalloc(alloc_size, GFP_KERNEL);
5596
	if (!p)
L
Linus Torvalds 已提交
5597 5598
		return NULL;

5599
	dev = PTR_ALIGN(p, NETDEV_ALIGN);
L
Linus Torvalds 已提交
5600
	dev->padded = (char *)dev - (char *)p;
5601

E
Eric Dumazet 已提交
5602 5603
	dev->pcpu_refcnt = alloc_percpu(int);
	if (!dev->pcpu_refcnt)
5604
		goto free_p;
5605 5606

	if (dev_addr_init(dev))
E
Eric Dumazet 已提交
5607
		goto free_pcpu;
5608

5609
	dev_mc_init(dev);
5610
	dev_uc_init(dev);
J
Jiri Pirko 已提交
5611

5612
	dev_net_set(dev, &init_net);
L
Linus Torvalds 已提交
5613

5614
	dev->gso_max_size = GSO_MAX_SIZE;
5615
	dev->gso_max_segs = GSO_MAX_SEGS;
5616 5617 5618 5619

	INIT_LIST_HEAD(&dev->napi_list);
	INIT_LIST_HEAD(&dev->unreg_list);
	INIT_LIST_HEAD(&dev->link_watch_list);
J
Jiri Pirko 已提交
5620
	INIT_LIST_HEAD(&dev->upper_dev_list);
5621 5622 5623
	dev->priv_flags = IFF_XMIT_DST_RELEASE;
	setup(dev);

T
Tom Herbert 已提交
5624 5625
	dev->num_tx_queues = txqs;
	dev->real_num_tx_queues = txqs;
5626
	if (netif_alloc_netdev_queues(dev))
5627
		goto free_all;
5628

E
Eric Dumazet 已提交
5629
#ifdef CONFIG_RPS
T
Tom Herbert 已提交
5630 5631
	dev->num_rx_queues = rxqs;
	dev->real_num_rx_queues = rxqs;
T
Tom Herbert 已提交
5632
	if (netif_alloc_rx_queues(dev))
5633
		goto free_all;
E
Eric Dumazet 已提交
5634
#endif
T
Tom Herbert 已提交
5635

L
Linus Torvalds 已提交
5636
	strcpy(dev->name, name);
5637
	dev->group = INIT_NETDEV_GROUP;
5638 5639
	if (!dev->ethtool_ops)
		dev->ethtool_ops = &default_ethtool_ops;
L
Linus Torvalds 已提交
5640
	return dev;
5641

5642 5643 5644 5645
free_all:
	free_netdev(dev);
	return NULL;

E
Eric Dumazet 已提交
5646 5647
free_pcpu:
	free_percpu(dev->pcpu_refcnt);
5648
	kfree(dev->_tx);
T
Tom Herbert 已提交
5649 5650 5651 5652
#ifdef CONFIG_RPS
	kfree(dev->_rx);
#endif

5653 5654 5655
free_p:
	kfree(p);
	return NULL;
L
Linus Torvalds 已提交
5656
}
T
Tom Herbert 已提交
5657
EXPORT_SYMBOL(alloc_netdev_mqs);
L
Linus Torvalds 已提交
5658 5659 5660 5661 5662

/**
 *	free_netdev - free network device
 *	@dev: device
 *
5663 5664
 *	This function does the last stage of destroying an allocated device
 * 	interface. The reference to the device object is released.
L
Linus Torvalds 已提交
5665 5666 5667 5668
 *	If this is the last reference then it will be freed.
 */
void free_netdev(struct net_device *dev)
{
5669 5670
	struct napi_struct *p, *n;

5671 5672
	release_net(dev_net(dev));

5673
	kfree(dev->_tx);
T
Tom Herbert 已提交
5674 5675 5676
#ifdef CONFIG_RPS
	kfree(dev->_rx);
#endif
5677

5678
	kfree(rcu_dereference_protected(dev->ingress_queue, 1));
5679

5680 5681 5682
	/* Flush device addresses */
	dev_addr_flush(dev);

5683 5684 5685
	list_for_each_entry_safe(p, n, &dev->napi_list, dev_list)
		netif_napi_del(p);

E
Eric Dumazet 已提交
5686 5687 5688
	free_percpu(dev->pcpu_refcnt);
	dev->pcpu_refcnt = NULL;

S
Stephen Hemminger 已提交
5689
	/*  Compatibility with error handling in drivers */
L
Linus Torvalds 已提交
5690 5691 5692 5693 5694 5695 5696 5697
	if (dev->reg_state == NETREG_UNINITIALIZED) {
		kfree((char *)dev - dev->padded);
		return;
	}

	BUG_ON(dev->reg_state != NETREG_UNREGISTERED);
	dev->reg_state = NETREG_RELEASED;

5698 5699
	/* will free via device release */
	put_device(&dev->dev);
L
Linus Torvalds 已提交
5700
}
E
Eric Dumazet 已提交
5701
EXPORT_SYMBOL(free_netdev);
5702

5703 5704 5705 5706 5707 5708
/**
 *	synchronize_net -  Synchronize with packet receive processing
 *
 *	Wait for packets currently being received to be done.
 *	Does not block later packets from starting.
 */
5709
void synchronize_net(void)
L
Linus Torvalds 已提交
5710 5711
{
	might_sleep();
5712 5713 5714 5715
	if (rtnl_is_locked())
		synchronize_rcu_expedited();
	else
		synchronize_rcu();
L
Linus Torvalds 已提交
5716
}
E
Eric Dumazet 已提交
5717
EXPORT_SYMBOL(synchronize_net);
L
Linus Torvalds 已提交
5718 5719

/**
5720
 *	unregister_netdevice_queue - remove device from the kernel
L
Linus Torvalds 已提交
5721
 *	@dev: device
5722
 *	@head: list
5723
 *
L
Linus Torvalds 已提交
5724
 *	This function shuts down a device interface and removes it
5725
 *	from the kernel tables.
5726
 *	If head not NULL, device is queued to be unregistered later.
L
Linus Torvalds 已提交
5727 5728 5729 5730 5731
 *
 *	Callers must hold the rtnl semaphore.  You may want
 *	unregister_netdev() instead of this.
 */

5732
void unregister_netdevice_queue(struct net_device *dev, struct list_head *head)
L
Linus Torvalds 已提交
5733
{
5734 5735
	ASSERT_RTNL();

5736
	if (head) {
5737
		list_move_tail(&dev->unreg_list, head);
5738 5739 5740 5741 5742
	} else {
		rollback_registered(dev);
		/* Finish processing unregister after unlock */
		net_set_todo(dev);
	}
L
Linus Torvalds 已提交
5743
}
5744
EXPORT_SYMBOL(unregister_netdevice_queue);
L
Linus Torvalds 已提交
5745

5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759
/**
 *	unregister_netdevice_many - unregister many devices
 *	@head: list of devices
 */
void unregister_netdevice_many(struct list_head *head)
{
	struct net_device *dev;

	if (!list_empty(head)) {
		rollback_registered_many(head);
		list_for_each_entry(dev, head, unreg_list)
			net_set_todo(dev);
	}
}
5760
EXPORT_SYMBOL(unregister_netdevice_many);
5761

L
Linus Torvalds 已提交
5762 5763 5764 5765 5766
/**
 *	unregister_netdev - remove device from the kernel
 *	@dev: device
 *
 *	This function shuts down a device interface and removes it
5767
 *	from the kernel tables.
L
Linus Torvalds 已提交
5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779 5780
 *
 *	This is just a wrapper for unregister_netdevice that takes
 *	the rtnl semaphore.  In general you want to use this and not
 *	unregister_netdevice.
 */
void unregister_netdev(struct net_device *dev)
{
	rtnl_lock();
	unregister_netdevice(dev);
	rtnl_unlock();
}
EXPORT_SYMBOL(unregister_netdev);

5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803 5804 5805 5806 5807 5808 5809 5810 5811
/**
 *	dev_change_net_namespace - move device to different nethost namespace
 *	@dev: device
 *	@net: network namespace
 *	@pat: If not NULL name pattern to try if the current device name
 *	      is already taken in the destination network namespace.
 *
 *	This function shuts down a device interface and moves it
 *	to a new network namespace. On success 0 is returned, on
 *	a failure a netagive errno code is returned.
 *
 *	Callers must hold the rtnl semaphore.
 */

int dev_change_net_namespace(struct net_device *dev, struct net *net, const char *pat)
{
	int err;

	ASSERT_RTNL();

	/* Don't allow namespace local devices to be moved. */
	err = -EINVAL;
	if (dev->features & NETIF_F_NETNS_LOCAL)
		goto out;

	/* Ensure the device has been registrered */
	if (dev->reg_state != NETREG_REGISTERED)
		goto out;

	/* Get out if there is nothing todo */
	err = 0;
5812
	if (net_eq(dev_net(dev), net))
5813 5814 5815 5816 5817 5818
		goto out;

	/* Pick the destination device name, and ensure
	 * we can use it in the destination network namespace.
	 */
	err = -EEXIST;
5819
	if (__dev_get_by_name(net, dev->name)) {
5820 5821 5822
		/* We get here if we can't use the current device name */
		if (!pat)
			goto out;
5823
		if (dev_get_valid_name(net, dev, pat) < 0)
5824 5825 5826 5827 5828 5829 5830 5831
			goto out;
	}

	/*
	 * And now a mini version of register_netdevice unregister_netdevice.
	 */

	/* If device is running close it first. */
5832
	dev_close(dev);
5833 5834 5835 5836 5837 5838 5839 5840 5841 5842 5843 5844

	/* And unlink it from device chain */
	err = -ENODEV;
	unlist_netdevice(dev);

	synchronize_net();

	/* Shutdown queueing discipline. */
	dev_shutdown(dev);

	/* Notify protocols, that we are about to destroy
	   this device. They should clean all the things.
5845 5846 5847 5848

	   Note that dev->reg_state stays at NETREG_REGISTERED.
	   This is wanted because this way 8021q and macvlan know
	   the device is just moving and can keep their slaves up.
5849 5850
	*/
	call_netdevice_notifiers(NETDEV_UNREGISTER, dev);
5851 5852
	rcu_barrier();
	call_netdevice_notifiers(NETDEV_UNREGISTER_FINAL, dev);
5853
	rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
5854 5855 5856 5857

	/*
	 *	Flush the unicast and multicast chains
	 */
5858
	dev_uc_flush(dev);
5859
	dev_mc_flush(dev);
5860

5861 5862 5863
	/* Send a netdev-removed uevent to the old namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_REMOVE);

5864
	/* Actually switch the network namespace */
5865
	dev_net_set(dev, net);
5866 5867 5868 5869 5870 5871 5872 5873 5874

	/* If there is an ifindex conflict assign a new one */
	if (__dev_get_by_index(net, dev->ifindex)) {
		int iflink = (dev->iflink == dev->ifindex);
		dev->ifindex = dev_new_index(net);
		if (iflink)
			dev->iflink = dev->ifindex;
	}

5875 5876 5877
	/* Send a netdev-add uevent to the new namespace */
	kobject_uevent(&dev->dev.kobj, KOBJ_ADD);

5878
	/* Fixup kobjects */
5879
	err = device_rename(&dev->dev, dev->name);
5880
	WARN_ON(err);
5881 5882 5883 5884 5885 5886 5887

	/* Add the device back in the hashes */
	list_netdevice(dev);

	/* Notify protocols, that a new device appeared. */
	call_netdevice_notifiers(NETDEV_REGISTER, dev);

5888 5889 5890 5891 5892 5893
	/*
	 *	Prevent userspace races by waiting until the network
	 *	device is fully setup before sending notifications.
	 */
	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);

5894 5895 5896 5897 5898
	synchronize_net();
	err = 0;
out:
	return err;
}
5899
EXPORT_SYMBOL_GPL(dev_change_net_namespace);
5900

L
Linus Torvalds 已提交
5901 5902 5903 5904 5905 5906 5907 5908 5909
static int dev_cpu_callback(struct notifier_block *nfb,
			    unsigned long action,
			    void *ocpu)
{
	struct sk_buff **list_skb;
	struct sk_buff *skb;
	unsigned int cpu, oldcpu = (unsigned long)ocpu;
	struct softnet_data *sd, *oldsd;

5910
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
L
Linus Torvalds 已提交
5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926
		return NOTIFY_OK;

	local_irq_disable();
	cpu = smp_processor_id();
	sd = &per_cpu(softnet_data, cpu);
	oldsd = &per_cpu(softnet_data, oldcpu);

	/* Find end of our completion_queue. */
	list_skb = &sd->completion_queue;
	while (*list_skb)
		list_skb = &(*list_skb)->next;
	/* Append completion queue from offline CPU. */
	*list_skb = oldsd->completion_queue;
	oldsd->completion_queue = NULL;

	/* Append output queue from offline CPU. */
5927 5928 5929 5930 5931 5932
	if (oldsd->output_queue) {
		*sd->output_queue_tailp = oldsd->output_queue;
		sd->output_queue_tailp = oldsd->output_queue_tailp;
		oldsd->output_queue = NULL;
		oldsd->output_queue_tailp = &oldsd->output_queue;
	}
5933 5934 5935 5936 5937
	/* Append NAPI poll list from offline CPU. */
	if (!list_empty(&oldsd->poll_list)) {
		list_splice_init(&oldsd->poll_list, &sd->poll_list);
		raise_softirq_irqoff(NET_RX_SOFTIRQ);
	}
L
Linus Torvalds 已提交
5938 5939 5940 5941 5942

	raise_softirq_irqoff(NET_TX_SOFTIRQ);
	local_irq_enable();

	/* Process offline CPU's input_pkt_queue */
5943
	while ((skb = __skb_dequeue(&oldsd->process_queue))) {
L
Linus Torvalds 已提交
5944
		netif_rx(skb);
5945
		input_queue_head_incr(oldsd);
T
Tom Herbert 已提交
5946
	}
5947
	while ((skb = __skb_dequeue(&oldsd->input_pkt_queue))) {
5948
		netif_rx(skb);
5949 5950
		input_queue_head_incr(oldsd);
	}
L
Linus Torvalds 已提交
5951 5952 5953 5954 5955

	return NOTIFY_OK;
}


5956
/**
5957 5958 5959 5960
 *	netdev_increment_features - increment feature set by one
 *	@all: current feature set
 *	@one: new feature set
 *	@mask: mask feature set
5961 5962
 *
 *	Computes a new feature set after adding a device with feature set
5963 5964
 *	@one to the master device with current feature set @all.  Will not
 *	enable anything that is off in @mask. Returns the new feature set.
5965
 */
5966 5967
netdev_features_t netdev_increment_features(netdev_features_t all,
	netdev_features_t one, netdev_features_t mask)
5968
{
5969 5970 5971
	if (mask & NETIF_F_GEN_CSUM)
		mask |= NETIF_F_ALL_CSUM;
	mask |= NETIF_F_VLAN_CHALLENGED;
5972

5973 5974
	all |= one & (NETIF_F_ONE_FOR_ALL|NETIF_F_ALL_CSUM) & mask;
	all &= one | ~NETIF_F_ALL_FOR_ALL;
5975

5976 5977 5978
	/* If one device supports hw checksumming, set for all. */
	if (all & NETIF_F_GEN_CSUM)
		all &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_GEN_CSUM);
5979 5980 5981

	return all;
}
5982
EXPORT_SYMBOL(netdev_increment_features);
5983

5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996
static struct hlist_head *netdev_create_hash(void)
{
	int i;
	struct hlist_head *hash;

	hash = kmalloc(sizeof(*hash) * NETDEV_HASHENTRIES, GFP_KERNEL);
	if (hash != NULL)
		for (i = 0; i < NETDEV_HASHENTRIES; i++)
			INIT_HLIST_HEAD(&hash[i]);

	return hash;
}

5997
/* Initialize per network namespace state */
5998
static int __net_init netdev_init(struct net *net)
5999
{
6000 6001
	if (net != &init_net)
		INIT_LIST_HEAD(&net->dev_base_head);
6002

6003 6004 6005
	net->dev_name_head = netdev_create_hash();
	if (net->dev_name_head == NULL)
		goto err_name;
6006

6007 6008 6009
	net->dev_index_head = netdev_create_hash();
	if (net->dev_index_head == NULL)
		goto err_idx;
6010 6011

	return 0;
6012 6013 6014 6015 6016

err_idx:
	kfree(net->dev_name_head);
err_name:
	return -ENOMEM;
6017 6018
}

6019 6020 6021 6022 6023 6024
/**
 *	netdev_drivername - network driver for the device
 *	@dev: network device
 *
 *	Determine network driver for device.
 */
6025
const char *netdev_drivername(const struct net_device *dev)
6026
{
6027 6028
	const struct device_driver *driver;
	const struct device *parent;
6029
	const char *empty = "";
6030 6031 6032

	parent = dev->dev.parent;
	if (!parent)
6033
		return empty;
6034 6035 6036

	driver = parent->driver;
	if (driver && driver->name)
6037 6038
		return driver->name;
	return empty;
6039 6040
}

6041
static int __netdev_printk(const char *level, const struct net_device *dev,
6042 6043 6044 6045
			   struct va_format *vaf)
{
	int r;

6046
	if (dev && dev->dev.parent) {
6047 6048 6049 6050 6051 6052
		r = dev_printk_emit(level[1] - '0',
				    dev->dev.parent,
				    "%s %s %s: %pV",
				    dev_driver_string(dev->dev.parent),
				    dev_name(dev->dev.parent),
				    netdev_name(dev), vaf);
6053
	} else if (dev) {
6054
		r = printk("%s%s: %pV", level, netdev_name(dev), vaf);
6055
	} else {
6056
		r = printk("%s(NULL net_device): %pV", level, vaf);
6057
	}
6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074

	return r;
}

int netdev_printk(const char *level, const struct net_device *dev,
		  const char *format, ...)
{
	struct va_format vaf;
	va_list args;
	int r;

	va_start(args, format);

	vaf.fmt = format;
	vaf.va = &args;

	r = __netdev_printk(level, dev, &vaf);
6075

6076 6077 6078 6079 6080 6081 6082 6083 6084 6085 6086 6087 6088 6089 6090 6091 6092 6093 6094
	va_end(args);

	return r;
}
EXPORT_SYMBOL(netdev_printk);

#define define_netdev_printk_level(func, level)			\
int func(const struct net_device *dev, const char *fmt, ...)	\
{								\
	int r;							\
	struct va_format vaf;					\
	va_list args;						\
								\
	va_start(args, fmt);					\
								\
	vaf.fmt = fmt;						\
	vaf.va = &args;						\
								\
	r = __netdev_printk(level, dev, &vaf);			\
6095
								\
6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109
	va_end(args);						\
								\
	return r;						\
}								\
EXPORT_SYMBOL(func);

define_netdev_printk_level(netdev_emerg, KERN_EMERG);
define_netdev_printk_level(netdev_alert, KERN_ALERT);
define_netdev_printk_level(netdev_crit, KERN_CRIT);
define_netdev_printk_level(netdev_err, KERN_ERR);
define_netdev_printk_level(netdev_warn, KERN_WARNING);
define_netdev_printk_level(netdev_notice, KERN_NOTICE);
define_netdev_printk_level(netdev_info, KERN_INFO);

6110
static void __net_exit netdev_exit(struct net *net)
6111 6112 6113 6114 6115
{
	kfree(net->dev_name_head);
	kfree(net->dev_index_head);
}

6116
static struct pernet_operations __net_initdata netdev_net_ops = {
6117 6118 6119 6120
	.init = netdev_init,
	.exit = netdev_exit,
};

6121
static void __net_exit default_device_exit(struct net *net)
6122
{
6123
	struct net_device *dev, *aux;
6124
	/*
6125
	 * Push all migratable network devices back to the
6126 6127 6128
	 * initial network namespace
	 */
	rtnl_lock();
6129
	for_each_netdev_safe(net, dev, aux) {
6130
		int err;
6131
		char fb_name[IFNAMSIZ];
6132 6133 6134 6135 6136

		/* Ignore unmoveable devices (i.e. loopback) */
		if (dev->features & NETIF_F_NETNS_LOCAL)
			continue;

6137 6138 6139
		/* Leave virtual devices for the generic cleanup */
		if (dev->rtnl_link_ops)
			continue;
6140

L
Lucas De Marchi 已提交
6141
		/* Push remaining network devices to init_net */
6142 6143
		snprintf(fb_name, IFNAMSIZ, "dev%d", dev->ifindex);
		err = dev_change_net_namespace(dev, &init_net, fb_name);
6144
		if (err) {
6145 6146
			pr_emerg("%s: failed to move %s to init_net: %d\n",
				 __func__, dev->name, err);
6147
			BUG();
6148 6149 6150 6151 6152
		}
	}
	rtnl_unlock();
}

6153 6154 6155
static void __net_exit default_device_exit_batch(struct list_head *net_list)
{
	/* At exit all network devices most be removed from a network
6156
	 * namespace.  Do this in the reverse order of registration.
6157 6158 6159 6160 6161 6162 6163 6164 6165 6166 6167 6168 6169 6170 6171 6172 6173
	 * Do this across as many network namespaces as possible to
	 * improve batching efficiency.
	 */
	struct net_device *dev;
	struct net *net;
	LIST_HEAD(dev_kill_list);

	rtnl_lock();
	list_for_each_entry(net, net_list, exit_list) {
		for_each_netdev_reverse(net, dev) {
			if (dev->rtnl_link_ops)
				dev->rtnl_link_ops->dellink(dev, &dev_kill_list);
			else
				unregister_netdevice_queue(dev, &dev_kill_list);
		}
	}
	unregister_netdevice_many(&dev_kill_list);
E
Eric Dumazet 已提交
6174
	list_del(&dev_kill_list);
6175 6176 6177
	rtnl_unlock();
}

6178
static struct pernet_operations __net_initdata default_device_ops = {
6179
	.exit = default_device_exit,
6180
	.exit_batch = default_device_exit_batch,
6181 6182
};

L
Linus Torvalds 已提交
6183 6184 6185 6186 6187 6188 6189 6190 6191 6192 6193 6194 6195 6196 6197 6198 6199 6200 6201 6202
/*
 *	Initialize the DEV module. At boot time this walks the device list and
 *	unhooks any devices that fail to initialise (normally hardware not
 *	present) and leaves us with a valid list of present and active devices.
 *
 */

/*
 *       This is called single threaded during boot, so no need
 *       to take the rtnl semaphore.
 */
static int __init net_dev_init(void)
{
	int i, rc = -ENOMEM;

	BUG_ON(!dev_boot_phase);

	if (dev_proc_init())
		goto out;

6203
	if (netdev_kobject_init())
L
Linus Torvalds 已提交
6204 6205 6206
		goto out;

	INIT_LIST_HEAD(&ptype_all);
6207
	for (i = 0; i < PTYPE_HASH_SIZE; i++)
L
Linus Torvalds 已提交
6208 6209
		INIT_LIST_HEAD(&ptype_base[i]);

6210 6211
	INIT_LIST_HEAD(&offload_base);

6212 6213
	if (register_pernet_subsys(&netdev_net_ops))
		goto out;
L
Linus Torvalds 已提交
6214 6215 6216 6217 6218

	/*
	 *	Initialise the packet receive queues.
	 */

6219
	for_each_possible_cpu(i) {
E
Eric Dumazet 已提交
6220
		struct softnet_data *sd = &per_cpu(softnet_data, i);
L
Linus Torvalds 已提交
6221

C
Changli Gao 已提交
6222
		memset(sd, 0, sizeof(*sd));
E
Eric Dumazet 已提交
6223
		skb_queue_head_init(&sd->input_pkt_queue);
6224
		skb_queue_head_init(&sd->process_queue);
E
Eric Dumazet 已提交
6225 6226
		sd->completion_queue = NULL;
		INIT_LIST_HEAD(&sd->poll_list);
6227 6228
		sd->output_queue = NULL;
		sd->output_queue_tailp = &sd->output_queue;
E
Eric Dumazet 已提交
6229
#ifdef CONFIG_RPS
E
Eric Dumazet 已提交
6230 6231 6232 6233
		sd->csd.func = rps_trigger_softirq;
		sd->csd.info = sd;
		sd->csd.flags = 0;
		sd->cpu = i;
6234
#endif
T
Tom Herbert 已提交
6235

E
Eric Dumazet 已提交
6236 6237 6238 6239
		sd->backlog.poll = process_backlog;
		sd->backlog.weight = weight_p;
		sd->backlog.gro_list = NULL;
		sd->backlog.gro_count = 0;
L
Linus Torvalds 已提交
6240 6241 6242 6243
	}

	dev_boot_phase = 0;

6244 6245 6246 6247 6248 6249 6250 6251 6252 6253 6254 6255 6256 6257 6258
	/* The loopback device is special if any other network devices
	 * is present in a network namespace the loopback device must
	 * be present. Since we now dynamically allocate and free the
	 * loopback device ensure this invariant is maintained by
	 * keeping the loopback device as the first device on the
	 * list of network devices.  Ensuring the loopback devices
	 * is the first device that appears and the last network device
	 * that disappears.
	 */
	if (register_pernet_device(&loopback_net_ops))
		goto out;

	if (register_pernet_device(&default_device_ops))
		goto out;

6259 6260
	open_softirq(NET_TX_SOFTIRQ, net_tx_action);
	open_softirq(NET_RX_SOFTIRQ, net_rx_action);
L
Linus Torvalds 已提交
6261 6262 6263 6264 6265 6266 6267 6268 6269

	hotcpu_notifier(dev_cpu_callback, 0);
	dst_init();
	rc = 0;
out:
	return rc;
}

subsys_initcall(net_dev_init);